From Koen.Kerremans@ehb.be Fri Jan 17 10:14:53 2003 From: Koen.Kerremans@ehb.be (KERREMANS, Koen) Date: Fri, 17 Jan 2003 11:14:53 +0100 Subject: [MT-List] multilingual terminography: towards intelligent dictionaries? Message-ID: <000001c2be11$47ce9e80$fc09a9c0@studttk.ehb.be> Multilingual Terminography: Towards Intelligent Dictionaries 28th of February in Brussels The Department of Applied Linguistics of the Erasmushogeschool Brussel would like to bring to your attention its upcoming seminar on 'Multilingual Terminography: Towards Intelligent Dictionaries'. This seminar, which will take place on the 28th of February and will have Jeanne Dancette as keynote speaker, shall focus on topics concerning: - 'terminography and users analysis' - 'semantic networks and ontologies'. More info can be found on the following website: http://ltc.ehb.be/terminography/ Kind regards, Koen Kerremans Research Assistent Department of Applied Linguistics Erasmushogeschool Brussel Trierstraat 84 B-1040 Brussels, Belgium koen.kerremans@ehb.be http://ttk.ehb.be From debe@comp.leeds.ac.uk Tue Jan 21 13:53:59 2003 From: debe@comp.leeds.ac.uk (D Elliott) Date: Tue, 21 Jan 2003 13:53:59 +0000 (GMT) Subject: [MT-List] Attn: Users of MT systems Message-ID: (Apologies for multiple postings.) Dear All I hope that the following will be of interest to MT users throughout the world. As part of my PhD research in the School of Computing and the Centre for Translation Studies at the University of Leeds, England, I will shortly begin compiling a multilingual corpus, comprising source texts aligned with human and machine translations, to be used for the evaluation of MT systems. The text types, topics, source and target languages for inclusion in the corpus will be based on information gathered from this short questionnaire (there are only 4 questions to answer!). Corpus content should, therefore, truly represent the kinds of texts that translation departments and companies translate most often with the aid of fully automatic MT systems. ALL RESPONDENTS WILL BE GRANTED FREE ACCESS TO THE CORPUS AND TO OUR MT EVALUATION RESULTS AS SOON AS THEY BECOME AVAILABLE. The information you provide is strictly confidential and will be used solely for the purpose of research into corpus design. Company data or individual information will not be disclosed to any external third parties. Please disregard any questions you are unwilling or unable to answer. We thank you in advance for your time. ________________________________________________________________________ MT USER SURVEY Name of Company: Nature of Business: Contact Name and Email: Role within Company: _________________________________________________________________________ Q1/4 Which fully automatic MT system(s) do you use? Please delete as appropriate. SDL Enterprise Translation Server Linguatec Personal Translator LogoMedia Enterprise Solutions Logo Vista X Multilingual/Pro MLTS Reverso Expert/Mac/Pro Smart Translator Systran Enterprise/Pro SDL Transcend Corporate Other (please specify) _________________________________________________________________________ Q2/4 Which language pair(s) do you translate using MT? (Please list below) Source language(s): Target language(s): _________________________________________________________________________ Q3/4 Using the following list as a guide, please give details of the text types you translate with the aid of fully automatic MT systems. For each text type you select, please give the approximate monthly word count if possible (or the percentage of the total amount of words machine translated) and details of specific subject areas if applicable. This information will be most helpful for our corpus content. Emails Memoranda Business letters Internal Company documents (eg. policy documents, training materials) Tourist/Travel information Web pages Newspaper articles Academic papers Calls for tender Patents Legal documents Scientific documents Technical documents Medical documents Instruction booklets User manuals Other (please specify) _________________________________________________________________________ Q4/4 For what purpose is the MT output used? (Please delete as appropriate) Gisting Immediate circulation Post-editing for internal use Post-editing for publication or client _________________________________________________________________________ THANK YOU FOR TAKING THE TIME TO COMPLETE THIS SURVEY. Please do not hesitate to contact me if you would like further information about this project. *************************************************** Debbie Elliott Centre for the Computer Analysis of Language and Speech, School of Computing, University of Leeds, Leeds LS2 9JT, United Kingdom. Tel: 0113 3436818 Email: debe@comp.leeds.ac.uk *************************************************** From evelyn@nats.informatik.uni-hamburg.de Fri Jan 17 14:47:45 2003 From: evelyn@nats.informatik.uni-hamburg.de (Evelyn Gius) Date: Fri, 17 Jan 2003 15:47:45 +0100 Subject: [MT-List] information about german machine readable lexicons Message-ID: <200301171547.45115.evelyn@nats15> Dear members of MT-list, I am a student of german language and literature and informatics at the University of Hamburg and I'm writing my thesis about machine readable german lexicons. I'm planning to make a comparison between the different lexicon formats in use. For this purpose I'm collecting the following information about the lexicons: - the lexicon scheme (= parts of the entry and their possible values), and - some entry samples which may be looking like the folling one: tab ; ; < > ... return values for : Nom v Adj values for : 1...13, for the classes 1 = ..., 2= ..., 3 = ...; etc. If you are - or have been - working with machine readable german lexicons of any type and size I would like to kindly ask you to provide me some information about those lexicons. (As I said above: I'm interested only in the formal aspects of the lexicons therefore I will neither evaluate nor use them!) Thank you for your cooperation, with regards, Evelyn Gius ------- Evelyn Gius Weidenallee 30d 20357 Hamburg phone: +49/ 40/ 45 000 682 email: evelyn@nats.informatik.uni-hamburg.de From wiggjd@sbu.ac.uk Thu Jan 30 19:14:26 2003 From: wiggjd@sbu.ac.uk (David Wigg) Date: Thu, 30 Jan 2003 19:14:26 +0000 Subject: [MT-List] Accented characters in e-mails Message-ID: <3E397992.BFAA6A03@sbu.ac.uk> This is a multi-part message in MIME format. --------------AAAA09CD11DCC252FCDAF152 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit I have been rather surprised by the number of e-mails I have received from some French and German sources in which normally accented characters have appeared incorrectly as other characters from the ASCII character set. This is almost always because an 8 bit character has been reduced to a 7 bit character by the loss of the left-most (senior) bit. I believe that the simplest, if not the only, solution at the current time to this problem is for SENDERS to set their e-mail systems to use MIME encoding with Quoted Printable set. This can in fact be left as the standard default whether using plain text or HTML. I attach a paper (in virus proof .rtf coding) for more information including how to do this for most popular e-mail systems. If anyone can inform me how to use MIME encoding with Quoted Printable set in any other e-mail systems please could they let me know so I can include this information in further issues of my paper. If anyone believes this advice and/or anything in the attached paper is unnecessary or incorrect I would be grateful if they would let me know. Thankyou. Yours sincerely, David Wigg Chairman Natural Language Translation Specilaist Group The British Computer Society --------------AAAA09CD11DCC252FCDAF152 Content-Type: application/rtf; name="AccentedCharacters2.rtf" Content-Transfer-Encoding: base64 Content-Disposition: inline; filename="AccentedCharacters2.rtf" e1xydGYxXGFuc2kgXGRlZmY0XGRlZmxhbmcxMDMze1xmb250dGJse1xmNFxmcm9tYW5cZmNo YXJzZXQwXGZwcnEyIFRpbWVzIE5ldyBSb21hbjt9e1xmMTExXGZtb2Rlcm5cZmNoYXJzZXQw XGZwcnExIENvdXJpZXIgKFcxKTt9fXtcY29sb3J0Ymw7XHJlZDBcZ3JlZW4wXGJsdWUwO1xy ZWQwXGdyZWVuMFxibHVlMjU1O1xyZWQwXGdyZWVuMjU1XGJsdWUyNTU7XHJlZDBcZ3JlZW4y NTVcYmx1ZTA7DQpccmVkMjU1XGdyZWVuMFxibHVlMjU1O1xyZWQyNTVcZ3JlZW4wXGJsdWUw O1xyZWQyNTVcZ3JlZW4yNTVcYmx1ZTA7XHJlZDI1NVxncmVlbjI1NVxibHVlMjU1O1xyZWQw XGdyZWVuMFxibHVlMTI4O1xyZWQwXGdyZWVuMTI4XGJsdWUxMjg7XHJlZDBcZ3JlZW4xMjhc Ymx1ZTA7XHJlZDEyOFxncmVlbjBcYmx1ZTEyODtccmVkMTI4XGdyZWVuMFxibHVlMDtccmVk MTI4XGdyZWVuMTI4XGJsdWUwO1xyZWQxMjhcZ3JlZW4xMjhcYmx1ZTEyODsNClxyZWQxOTJc Z3JlZW4xOTJcYmx1ZTE5Mjt9e1xzdHlsZXNoZWV0e1xub3dpZGN0bHBhciBcZjQgXHNuZXh0 MCBOb3JtYWw7fXtcKlxjczEwIFxhZGRpdGl2ZSBEZWZhdWx0IFBhcmFncmFwaCBGb250O319 e1xpbmZve1x0aXRsZSBBY2NlbnRlZCBhbmQgc3BlY2lhbCBjaGFyYWN0ZXJzIGluIGUtbWFp bHN9e1xhdXRob3IgbWF0dGhldy5zYXJhaC5qZXNzb3B9e1xvcGVyYXRvciBEYXZpZCBXaWdn fQ0Ke1xjcmVhdGltXHlyMjAwMlxtbzhcZHkyMFxocjE5fXtccmV2dGltXHlyMjAwMlxtbzEy XGR5MThcaHIyMVxtaW4xNH17XHByaW50aW1ceXIyMDAyXG1vMTJcZHk1XGhyOVxtaW4yMX17 XHZlcnNpb24yfXtcZWRtaW5zNDN9e1xub2ZwYWdlczh9e1xub2Z3b3JkczMwNTZ9e1xub2Zj aGFyczE3NDIyfXtcKlxjb21wYW55IFNvZnR3YXJlIEFuYWx5c2lzIFN5c3RlbXN9e1x2ZXJu NTc0MzF9fQ0KXHBhcGVydzExOTA5XHBhcGVyaDE2ODM0XG1hcmdsMTQ0MFxtYXJncjE0NDBc bWFyZ3Q3MjBcbWFyZ2I3MjAgXGRlZnRhYjE0NDBcd2lkb3djdHJsXGZ0bmJqXGFlbmRkb2Nc YWZ0bm5hclxoeXBoY2FwczBcZm9ybXNoYWRlIFxmZXQwXHNlY3RkIFxsaW5leDBcaGVhZGVy eTcwNlxmb290ZXJ5NzA2XGNvbHN4NzA4XGVuZG5oZXJlIHtcKlxwbnNlY2x2bDFccG51Y3Jt XHBuc3RhcnQxXHBuaW5kZW50NzIwXHBuaGFuZ3tccG50eHRhIC59fQ0Ke1wqXHBuc2VjbHZs MlxwbnVjbHRyXHBuc3RhcnQxXHBuaW5kZW50NzIwXHBuaGFuZ3tccG50eHRhIC59fXtcKlxw bnNlY2x2bDNccG5kZWNccG5zdGFydDFccG5pbmRlbnQ3MjBccG5oYW5ne1xwbnR4dGEgLn19 e1wqXHBuc2VjbHZsNFxwbmxjbHRyXHBuc3RhcnQxXHBuaW5kZW50NzIwXHBuaGFuZ3tccG50 eHRhICl9fXtcKlxwbnNlY2x2bDVccG5kZWNccG5zdGFydDFccG5pbmRlbnQ3MjBccG5oYW5n e1xwbnR4dGIgKH17XHBudHh0YSApfX0NCntcKlxwbnNlY2x2bDZccG5sY2x0clxwbnN0YXJ0 MVxwbmluZGVudDcyMFxwbmhhbmd7XHBudHh0YiAofXtccG50eHRhICl9fXtcKlxwbnNlY2x2 bDdccG5sY3JtXHBuc3RhcnQxXHBuaW5kZW50NzIwXHBuaGFuZ3tccG50eHRiICh9e1xwbnR4 dGEgKX19e1wqXHBuc2VjbHZsOFxwbmxjbHRyXHBuc3RhcnQxXHBuaW5kZW50NzIwXHBuaGFu Z3tccG50eHRiICh9e1xwbnR4dGEgKX19e1wqXHBuc2VjbHZsOQ0KXHBubGNybVxwbnN0YXJ0 MVxwbmluZGVudDcyMFxwbmhhbmd7XHBudHh0YiAofXtccG50eHRhICl9fVxwYXJkXHBsYWlu IFx3aWRjdGxwYXIgXGY0IHtcYlxmMTExXGZzMzIgQWNjZW50ZWQgY2hhcmFjdGVycyBpbiB0 ZXh0IGJhc2VkIGUtbWFpbHMgDQpccGFyIH17XGYxMTFcZnMyMCANClxwYXIgfXtcYlxmMTEx XGZzMjggVGhlIFByb2JsZW0NClxwYXIgfXtcZjExMVxmczIwIA0KXHBhciBBIHBvdW5kIHNp Z24gaW4gYW4gZS1tYWlsIGFycml2ZXMgYXMgYSBoYXNoIHNpZ24gb3IgYW4gYWNjZW50ZWQg b3Igb3RoZXIgc3BlY2lhbCBjaGFyYWN0ZXIgaXMgcmVjZWl2ZWQgYXMgYSBkaWZmZXJlbnQg bGV0dGVyIG9yIGNoYXJhY3Rlci4NClxwYXIgDQpccGFyIH17XGJcZjExMVxmczI4IFRoZSBT b2x1dGlvbg0KXHBhciB9e1xmMTExXGZzMjAgDQpccGFyIFRoZSBTRU5ERVIgbXVzdCBzZXQg dXAgdGhlaXIgZS1tYWlsIHN5c3RlbSBjb3JyZWN0bHkgYXMgZm9sbG93cy4NClxwYXIgDQpc cGFyIFRvIHNlbmQgYWNjZW50ZWQgb3Igc29tZSBzcGVjaWFsIGNoYXJhY3RlcnMgKGluY2x1 ZGluZyB0aGUgcG91bmQgc2lnbikgaW4gdGV4dCBiYXNlZCBlLW1haWxzIHlvdSBuZWVkIHRv IHNldCB0aGUgZm9sbG93aW5nIG9wdGlvbnMgaW4geW91ciBlLW1haWwgc3lzdGVtLA0KXHBh ciANClxwYXIgMS4gU2V0IE1JTUUgZW5jb2Rpbmcgd2l0aCBRdW90ZWQgUHJpbnRhYmxlIHNl dCwNClxwYXIgMi4gc2VsZWN0IGEgQ2hhcmFjdGVyIFNldCBzdWNoIGFzIElTTyA4ODU5LTEg KG9yIElTTyA4ODU5LTE1KSBmb3IgV2VzdGVybiBFdXJvcGVhbiBsYW5ndWFnZXMuIFRoZXJl IGFyZSBtYW55IG90aGVycyBmb3Igb3RoZXIgbGFuZ3VhZ2VzLg0KXHBhciANClxwYXIgU2Vl IGJlbG93IGZvciBob3cgdG8gc2V0IHRoZXNlIGZvciBzb21lIG9mIHRoZSBtb3JlIHBvcHVs YXIgZS1tYWlsIHN5c3RlbXMuIE1JTUUgZW5jb2Rpbmcgd2l0aCBRdW90ZWQgUHJpbnRhYmxl IHNldCBjYW4gYW5kIHNob3VsZCBiZSBsZWZ0IHBlcm1hbmVudGx5IHNldCBhcyB0aGV5IHdp bGwgbm90IG5lZWQgdG8gYmUgY2hhbmdlZCBmb3IgYW55IG90aGVyIHR5cGUgb2YgZS1tYWls Lg0KXHBhciANClxwYXIgT25jZSBzZXQgdXAsIHlvdSBjb3VsZCBzZW5kIGEgdGVzdCBtZXNz YWdlIHRvIGEgZnJpZW5kIGF0IGFub3RoZXIgc2l0ZSB0byBjaGVjayB0aGF0IHRoZSBwb3Vu ZCBzaWduIGFuZC9vciBhY2NlbnRlZCBjaGFyYWN0ZXJzIGFyZSBiZWluZyByZWNlaXZlZCBj b3JyZWN0bHkuDQpccGFyIA0KXHBhciBTZWUgYmVsb3cgZm9yIGV4cGxhbmF0aW9ucyB3aXRo IGZ1cnRoZXIgZGV0YWlsZWQgaW5mb3JtYXRpb24gYXMgZm9sbG93cywNClxwYXIgDQpccGFy IDEuIFNldHRpbmdzIGZvciBzb21lIGUtbWFpbCBzeXN0ZW1zDQpccGFyIDIuIFdlc3Rlcm4g RXVyb3BlYW4gY2hhcmFjdGVyIHNldHMNClxwYXIgMy4gTUlNRSBlbmNvZGluZw0KXHBhciA0 LiBPdGhlciBjaGFyYWN0ZXIgc2V0cw0KXHBhciA1LiBDaGFyYWN0ZXIgZW50cnkNClxwYXIg Ni4gUmVjZWl2aW5nIGUtbWFpbHMNClxwYXIgNy4gU291cmNlIGNvZGUgb2YgZS1tYWlscw0K XHBhciA4LiBHbG9zc2FyeQ0KXHBhciANClxwYXIgfXtcYlxmMTExIDEuIFNldHRpbmdzIGZv ciBzb21lIGUtbWFpbCBzeXN0ZW1zDQpccGFyIH17XGYxMTFcZnMyMCANClxwYXIgRGlmZmVy ZW50IHZlcnNpb25zIG9mIHRoZSBzYW1lIGUtbWFpbCBzeXN0ZW0gbWF5IHVzZSBzbGlnaHRs eSBkaWZmZXJlbnQgc2V0dGluZ3MgdG8gYWNoaWV2ZSB0aGUgc2FtZSByZXN1bHQuIFRoZSBm b2xsb3dpbmcgc2V0dGluZ3MgKGZvciBQQyBzeXN0ZW1zKSBzaG91bGQgYmUgYWxyaWdodCBm b3IgcmVjZW50IHZlcnNpb25zICgyMDAyKSAoSWYgbm90LCBwbGVhc2UgbGV0IG1lIGtub3cp LiBJZiBzb21lb25lIHdpbGwgc2VuZCBtZSBzZXR0aW5nDQpzIGZvciBNYWNzIEkgd2lsbCBh ZGQgdGhlbSB0byB0aGUgbGlzdC4NClxwYXIgDQpccGFyIERvblxycXVvdGUgdCBmb3JnZXQs IG9uY2Ugc2V0IHRoZXJlIGlzIG5vIG5lZWQgdG8gY2hhbmdlIHRoZSBNSU1FIGVuY29kaW5n IHdpdGggUXVvdGVkIFByaW50YWJsZSBzZXR0aW5nIGFnYWluLCBldmVuIGlmIHlvdSBzb21l dGltZXMgdXNlIEhUTUwuDQpccGFyIA0KXHBhciB9e1xiXGYxMTFcZnMyMCBFdWRvcmENClxw YXIgfXtcZjExMVxmczIwIA0KXHBhciBDbGljayBcbGRibHF1b3RlIFRvb2xzXHJkYmxxdW90 ZSAgbWVudSwgXGxkYmxxdW90ZSBPcHRpb25zXHJkYmxxdW90ZSAgYW5kIFxsZGJscXVvdGUg Q29tcG9zaW5nIE1haWxccmRibHF1b3RlICwNClxwYXIgdGhlbiB0aWNrIFxsZGJscXVvdGUg TWF5IFVzZSBRdW90ZWQtUHJpbnRhYmxlXHJkYmxxdW90ZSAuDQpccGFyIFlvdSBjYW4gYWxz byBmaW5kIFxsZGJscXVvdGUgQXR0YWNobWVudHNccmRibHF1b3RlICBpbiBcbGRibHF1b3Rl IE9wdGlvbnNccmRibHF1b3RlICBhbmQNClxwYXIgdGhlbiBzZWxlY3QgXGxkYmxxdW90ZSBF bmNvZGluZyBNZXRob2RccmRibHF1b3RlICB0byBcbGRibHF1b3RlIE1JTUVccmRibHF1b3Rl IC4NClxwYXIgQ2hhcmFjdGVyIHNldCBjYW5ub3QgYmUgY2hvc2VuLiBJdCBpcyBmaXhlZCBm b3IgY291bnRyeSBvZiBvcmlnaW4gb25seS4NClxwYXIgKE5vdGUgdGhhdCB0aGlzIGRvZXMg bm90IGluIGl0c2VsZiBwcmV2ZW50IHNlbmRpbmcgY2hhcmFjdGVycyBmcm9tIGFub3RoZXIg Y2hhcmFjdGVyIHNldCwgb25seSB0aGF0IHRoZSByZWNpcGllbnQgd2lsbCBoYXZlIHRvIGFz c3VtZSB0aGUgYXBwcm9wcmlhdGUgY2hhcmFjdGVyIHNldCwgd2hpY2ggd291bGQgcHJvYmFi bHkgYmUgdGhlaXIgb3duIGFueXdheS4pDQpccGFyIA0KXHBhciANClxwYXIgDQpccGFyIH17 XGJcZjExMVxmczIwIE5ldHNjYXBlIGFuZCBNb3ppbGxhDQpccGFyIH17XGYxMTFcZnMyMCAN ClxwYXIgQ2xpY2sgXGxkYmxxdW90ZSBFZGl0XHJkYmxxdW90ZSAgbWVudSwgXGxkYmxxdW90 ZSBQcmVmZXJlbmNlc1xyZGJscXVvdGUgIGFuZCBkb3VibGUgY2xpY2sgXGxkYmxxdW90ZSBN YWlsICYgTmV3c2dyb3Vwc1xyZGJscXVvdGUgLA0KXHBhciBjbGljayBcbGRibHF1b3RlIE1l c3NhZ2VzXHJkYmxxdW90ZSAgb3IgXGxkYmxxdW90ZSBDb21wb3NpdGlvblxyZGJscXVvdGUg IGFuZCANClxwYXIgdGhlbiBzZWxlY3QgXGxkYmxxdW90ZSBVc2luZyB0aGUgXGxxdW90ZSBR dW90ZWQgUHJpbnRhYmxlXHJxdW90ZSAgTUlNRSBFbmNvZGluZ1xyZGJscXVvdGUgLg0KXHBh ciBGb3IgQ2hhcmFjdGVyIFNldCBzZWxlY3Rpb24sDQpccGFyIGNsaWNrIFxsZGJscXVvdGUg Vmlld1xyZGJscXVvdGUgIG1lbnUsIGFuZCBcbGRibHF1b3RlIENoYXJhY3RlciBTZXQvQ29k aW5nXHJkYmxxdW90ZSAsDQpccGFyIHRoZW4gdGljayByZXF1aXJlZCBDaGFyYWN0ZXIgU2V0 Lg0KXHBhciBXaGVuIHNlbmRpbmcgZS1tYWlscyBjaGVjayB0aGF0IHRoZSBjb3JyZWN0IGNo YXJhY3RlciBzZXQgaXMgYmVpbmcgdXNlZCBpbiB0aGUgc2FtZSB3YXkgYnV0IHVzaW5nIHRo ZSBcbGRibHF1b3RlIFZpZXdccmRibHF1b3RlICBtZW51IGluIHRoZSBuZXcgbWVzc2FnZSBw YW5lbC4NClxwYXIgDQpccGFyIH17XGJcZjExMVxmczIwIE9wZXJhDQpccGFyIH17XGYxMTFc ZnMyMCANClxwYXIgRG9lcyBub3QgYXBwZWFyIHBvc3NpYmxlIHRvIHNldCBcbGRibHF1b3Rl IFF1b3RlZCBQcmludGFibGVccmRibHF1b3RlICAoU28gY291bGQgbm90IGV2ZW4gc2VuZCBw b3VuZCBzaWduLCBcbGRibHF1b3RlIFwnYTNccmRibHF1b3RlICkuDQpccGFyIEZvciBDaGFy YWN0ZXIgU2V0IHNlbGVjdGlvbiwgDQpccGFyIGNsaWNrIFxsZGJscXVvdGUgRmlsZVxyZGJs cXVvdGUgIG1lbnUsIFxsZGJscXVvdGUgUHJlZmVyZW5jZXNccmRibHF1b3RlICwgXGxkYmxx dW90ZSBFLW1haWxccmRibHF1b3RlICBhbmQgXGxkYmxxdW90ZSBQcm9wZXJ0aWVzXHJkYmxx dW90ZSAuIFRoZW4gaW4gXGxkYmxxdW90ZSBFLW1haWwgYWNjb3VudCBwcm9wZXJ0aWVzXHJk YmxxdW90ZSAgc2VsZWN0IFxsZGJscXVvdGUgT3V0Z29pbmdccmRibHF1b3RlIA0KIHRhYiBh bmQgc2VsZWN0IHJlcXVpcmVkIGNoYXJhY3RlciBzZXQgaW4gXGxkYmxxdW90ZSBGb250IGFu ZCBlbmNvZGluZ1xyZGJscXVvdGUgIHNlY3Rpb24uDQpccGFyIERvZXMgbm90IGFwcGVhciBw b3NzaWJsZSB0byBzZWxlY3QgYSBjaGFyYWN0ZXIgc2V0IGZvciBpbmNvbWluZyBtZXNzYWdl cy4NClxwYXIgDQpccGFyIH17XGJcZjExMVxmczIwIE91dGxvb2sgRXhwcmVzcw0KXHBhciB9 e1xmMTExXGZzMjAgDQpccGFyIENsaWNrIFxsZGJscXVvdGUgVG9vbHNccmRibHF1b3RlICBt ZW51LCBcbGRibHF1b3RlIE9wdGlvbnNccmRibHF1b3RlICBhbmQgY2xpY2sgdGhlIFxsZGJs cXVvdGUgU2VuZFxyZGJscXVvdGUgIHRhYiBhbmQgDQpccGFyIHRoZW4gdW5kZXIgXGxkYmxx dW90ZSBNYWlsIFNlbmRpbmcgRm9ybWF0XHJkYmxxdW90ZSAgc2VsZWN0IFxsZGJscXVvdGUg UGxhaW4gVGV4dFxyZGJscXVvdGUgLA0KXHBhciB0aGVuIGNsaWNrIFxsZGJscXVvdGUgUGxh aW4gVGV4dCBTZXR0aW5nc1xyZGJscXVvdGUgIGFuZCBzZWxlY3QgXGxkYmxxdW90ZSBNSU1F XHJkYmxxdW90ZSAgYW5kIFxsZGJscXVvdGUgUXVvdGVkIFByaW50YWJsZVxyZGJscXVvdGUg Lg0KXHBhciAoSWYgeW91IGFsc28gbmVlZCB0byBzZW5kIEhUTUwsIHNldCB0aGUgY29ycmVz cG9uZGluZyBIVE1MIHNldHRpbmdzLikNClxwYXIgVW5kZXIgXGxkYmxxdW90ZSBOZXdzIFNl bmRpbmcgRm9ybWF0XHJkYmxxdW90ZSAgeW91IG1heSB3YW50IHRvIG1ha2UgdGhlIGNvcnJl c3BvbmRpbmcgc2V0dGluZ3MgYXMgZm9yIFBsYWluIFRleHQgKGFuZCBIVE1MKS4NClxwYXIg Rm9yIHNlbGVjdGlvbiBvZiB5b3VyIGRlZmF1bHQgQ2hhcmFjdGVyIFNldCBmb3Igc2VuZGlu ZyBtZXNzYWdlcywNClxwYXIgY2xpY2sgXGxkYmxxdW90ZSBJbnRlcm5hdGlvbmFsIFNldHRp bmdzXHJkYmxxdW90ZSAgaW4gdGhlIFxsZGJscXVvdGUgU2VuZFxyZGJscXVvdGUgIHRhYiBh bmQgdGhlbiBmb3IgXGxkYmxxdW90ZSBEZWZhdWx0IEVuY29kaW5nXHJkYmxxdW90ZSAgc2Vs ZWN0IHJlcXVpcmVkIGNoYXJhY3RlciBzZXQgKGUuZy4gV2VzdGVybiBFdXJvcGVhbiAoSVNP KSApLg0KXHBhciBGb3Igc2VsZWN0aW9uIG9mIHlvdXIgZGVmYXVsdCBDaGFyYWN0ZXIgU2V0 IGZvciByZWNlaXZpbmcgbWVzc2FnZXMsDQpccGFyIGNsaWNrIFxsZGJscXVvdGUgRm9udHNc cmRibHF1b3RlICBpbiB0aGUgXGxkYmxxdW90ZSBSZWFkXHJkYmxxdW90ZSAgdGFiIGFuZCB0 aGVuIHNlbGVjdCB0aGUgcmVxdWlyZWQgZm9udCBzZXR0aW5ncyAoZS5nLiBXZXN0ZXJuIElT TykgYW5kIGNsaWNrIFxsZGJscXVvdGUgU2V0IGFzIGRlZmF1bHRccmRibHF1b3RlIC4NClxw YXIgVG8gc2VsZWN0IGEgZGlmZmVyZW50IENoYXJhY3RlciBTZXQgZm9yIHJlYWRpbmcgYSBw YXJ0aWN1bGFyIG1lc3NhZ2UgY2xpY2sgdGhlIFxsZGJscXVvdGUgVmlld1xyZGJscXVvdGUg IG1lbnUgYW5kIFxsZGJscXVvdGUgRW5jb2RpbmdccmRibHF1b3RlICBhbmQgdGhlbiBzZWxl Y3QgdGhlIGFwcHJvcHJpYXRlIENoYXJhY3RlciBTZXQgb3IgZm9yIHNlbmRpbmcgYSBuZXcg bWVzc2FnZSBjbGljayB0aGUgXGxkYmxxdW90ZSBGb3JtYXQNClxyZGJscXVvdGUgIG1lbnUg b24gdGhlIFxsZGJscXVvdGUgTmV3IE1lc3NhZ2VccmRibHF1b3RlICB3aW5kb3cgYW5kIHRo ZW4gc2VsZWN0IHRoZSByZXF1aXJlZCBDaGFyYWN0ZXIgU2V0Lg0KXHBhciANClxwYXIgfXtc YlxmMTExXGZzMjAgUGVnYXN1cw0KXHBhciB9e1xmMTExXGZzMjAgDQpccGFyIENsaWNrIFxs ZGJscXVvdGUgVG9vbHNccmRibHF1b3RlICBtZW51LCBhbmQgXGxkYmxxdW90ZSBPcHRpb25z XHJkYmxxdW90ZSAsDQpccGFyIGNsaWNrIFxsZGJscXVvdGUgQWR2YW5jZWQgc2V0dGluZ3Nc cmRibHF1b3RlICBhbmQgbm90ZSBkZWZhdWx0IE1JTUUgY2hhcmFjdGVyIHNldCwgcHJvYmFi bHkgSVNPLTg4NTktMSwNClxwYXIgY2xpY2sgXGxkYmxxdW90ZSBNZXNzYWdlcyBhbmQgUmVw bGllc1xyZGJscXVvdGUgIGFuZCBzZWxlY3QgXGxkYmxxdW90ZSBVc2UgTUlNRSBmZWF0dXJl cyBvblxyZGJscXVvdGUgLA0KXHBhciBjbGljayBcbGRibHF1b3RlIFNlbmRpbmcgbWFpbFxy ZGJscXVvdGUgIGFuZCB9e1xmMTExXGZzMjBcdWwgZGV9e1xmMTExXGZzMjAgc2VsZWN0IFxs ZGJscXVvdGUgQWxsb3cgOC1iaXQgTUlNRSBtZXNzYWdlIGVuY29kaW5nXHJkYmxxdW90ZSAu DQpccGFyIE5vdGUgdGhhdCB0aG91Z2ggaXQgYXBwZWFycyBwb3NzaWJsZSB0byBjaGFuZ2Ug dGhlIG5hbWUgb2YgdGhlIGRlZmF1bHQgTUlNRSBjaGFyYWN0ZXIgc2V0IGFib3ZlIEkgaGF2 ZSBvbmx5IHN1Y2NlZWRlZCBpbiBzZW5kaW5nIGFjY2VudGVkIGNoYXJhY3RlcnMgd2l0aCBJ U08tODg1OS0xIHdoaWNoIHdhcyBhcyBzZXQgd2hlbiB0aGUgc3lzdGVtIHdhcyBkb3dubG9h ZGVkLihOb3RlLiBKYXJvLkxham92aWNAbWYudW5pLWxqLnNpIGhhcyBiZWVuIA0KYWJsZSB0 byB1c2UgSVNPIDg4NTktMiBhbmQgaXMgd2lsbGluZyB0byBnaXZlIGFkdmljZSBvbiBob3cg dG8gZG8gdGhpcy4pIEl0IGRvZXMgbm90IGFwcGVhciBwb3NzaWJsZSB0byBjaGFuZ2UgdGhl IGNoYXJhY3RlciBzZXQgZm9yIHJlYWRpbmcgbWVzc2FnZXMuDQpccGFyIA0KXHBhciANClxw YXIgfXtcYlxmMTExIDIuIFdlc3Rlcm4gRXVyb3BlYW4gY2hhcmFjdGVyIHNldA0KXHBhciB9 e1xmMTExXGZzMjAgDQpccGFyIFRoZXJlIHNlZW1zIHRvIGJlIGEgc3VycHJpc2luZyBtaXN1 bmRlcnN0YW5kaW5nIGFuZCBsYWNrIG9mIGtubw0Kd2xlZGdlIGFib3V0IHRoZSB1c2Ugb2Yg YWNjZW50ZWQgYW5kIHNwZWNpYWwgY2hhcmFjdGVycyBpbiBlLW1haWxzLiBUaGUgaGVscCBm aWxlcyBvZiB0aGUgdmFyaW91cyBlLW1haWwgc3lzdGVtcyBhcmUgbm90IHVzdWFsbHkgdmVy eSBlbmxpZ2h0ZW5pbmcgYWJvdXQgaG93IHRvIHNlbmQgdGhlbSBlaXRoZXIuIFRoaXMgYXJ0 aWNsZSBzZXRzIG91dCB0byBleHBsYWluIHRoZSBwcm9ibGVtIGFuZCB0byBzaG93IGhvdyB5 b3VyIHN5c3RlbSBjYW4gDQpiZSBzZXQgdXAgZWFzaWx5IGFuZCBwZXJtYW5lbnRseSB0byBz ZW5kIGFjY2VudGVkIGFuZCBzcGVjaWFsIGNoYXJhY3RlcnMgdXNpbmcgdGhlIHZhc3QgbWFq b3JpdHkgb2YgY3VycmVudCBlLW1haWwgc3lzdGVtcy4NClxwYXIgDQpccGFyIEFsbCBjaGFy YWN0ZXJzIChpbmNsdWRpbmcgdGhlIGNoYXJhY3RlcnMgZm9yIHRoZSBkaWdpdHMpIGFyZSBy ZXByZXNlbnRlZCBpbiBjb21wdXRlcnMgYnkgbnVtYmVycyBmcm9tIDAgdG8gMjU1ICgyNTYg aW4gYWxsKS4gVGhpcyBiZWNhdXNlIHRoZXkgYXJlIGVhY2ggcmVwcmVzZW50ZWQgYnkgOCBi aW5hcnkgYml0cyBjYWxsZWQgYSBieXRlIHdoaWNoIGNhbiBzdG9yZSBhIHZhbHVlIG9mIDIg cmFpc2VkIHRvIHRoZSBwb3dlciA4ICgyNTYpLg0KDQpccGFyIA0KXHBhciBIb3dldmVyLCB0 aGUgSW50ZXJuZXQgaXMgKDIwMDIpIGxhcmdlbHkgYmFzZWQgb24gdGhlIHVzZSBvZiB0aGUg VVMgQVNDSUkgY2hhcmFjdGVyIHNldCBvZiBvbmx5IDEyOCBjaGFyYWN0ZXJzLCBhbmQgdGhl cmVpbiBsaWVzIHRoZSBtYWluIHByb2JsZW0gd2UgYXJlIGFkZHJlc3NpbmcuIFRoZSBmaXJz dCAzMiBvZiB0aGVzZSBhcmUgdmFyaW91cyBjb250cm9sIGNoYXJhY3RlcnMgb2Ygd2hpY2gg b25seSBvbmUsIG5hbWVseSB0aGUgVEFCIGNoYXINCmFjdGVyLCBpcyB1c2VkIG5vcm1hbGx5 IGJ5IHVzIGluIGUtbWFpbHMuIFRoaXMgbGVhdmVzIG9ubHkgOTYgY2hhcmFjdGVycyB3aGlj aCBjYW4gYmUgdXNlZCBmb3IgZS1tYWlscyAoMTAgZGlnaXRzLCAyNiBsb3dlciBjYXNlIGFu ZCAyNg0KIHVwcGVyIGNhc2UgbGV0dGVycyBhbmQgMzQgcHVuY3R1YXRpb24gYW5kIHNwZWNp YWwgY2hhcmFjdGVycykuIFRoZXNlIGFyZSB0aGUgb25seSBjaGFyYWN0ZXJzLCBvciByYXRo ZXIsIHZhbHVlcyByZXByZXNlbnRpbmcgY2hhcmFjdGVycywgd2hpY2ggY2FuIGJlIHNlbnQg dGhyb3VnaCB0aGUgaW50ZXJuZXQgKHNvIG9idmlvdXNseSB3ZSBuZWVkIHNvbWUgd2F5IG9m IHJlcHJlc2VudGluZyBhbGwgdGhlIG90aGVyIGNoYXJhY3RlcnMgd2UgbWlnDQpodCBuZWVk LCB3aGljaCB3ZSB3aWxsIGRpc2N1c3Mgc2hvcnRseSkuDQpccGFyIA0KXHBhciBUaGUgbnVt YmVycyBmcm9tIDAgdG8gMzEgYXJlIHVzZWQgZm9yIHZhcmlvdXMgY29udHJvbCBjaGFyYWN0 ZXJzLCBhbmQgdGhlIG51bWJlcnMgZnJvbSAzMiB0byAxMjcgcmVwcmVzZW50IHRoZSA5NiBj aGFyYWN0ZXJzIHdlIGNhbiBub3JtYWxseSB1c2UgZm9yIGUtbWFpbHMuIFRoZXNlIGFyZSBz aG93biBiZWxvdyB3aXRoIHRoZWlyIG51bWVyaWMgdmFsdWVzLg0KXHBhciANClxwYXIgMzIg IFx0YWIgNDggMFx0YWIgNjQgQFx0YWIgODAgUFx0YWIgOTYgIGBcdGFiIDExMiBwDQpccGFy IDMzICFcdGFiIDQ5IDFcdGFiIDY1IEFcdGFiIDgxIFFcdGFiIDk3ICBhXHRhYiAxMTMgcQ0K XHBhciAzNCBcbGRibHF1b3RlIFx0YWIgNTAgMlx0YWIgNjYgQlx0YWIgODIgUlx0YWIgOTgg IGJcdGFiIDExNCByDQpccGFyIDM1ICNcdGFiIDUxIDNcdGFiIDY3IENcdGFiIDgzIFNcdGFi IDk5ICBjXHRhYiAxMTUgcw0KXHBhciAzNiAkXHRhYiA1MiA0XHRhYiA2OCBEXHRhYiA4NCBU XHRhYiAxMDAgZFx0YWIgMTE2IHQNClxwYXIgMzcgJVx0YWIgNTMgNVx0YWIgNjkgRVx0YWIg ODUgVVx0YWIgMTAxIGVcdGFiIDExNyB1DQpccGFyIDM4ICZcdGFiIDU0IDZcdGFiIDcwIEZc dGFiIDg2IFZcdGFiIDEwMiBmXHRhYiAxMTggdg0KXHBhciAzOSBcbHF1b3RlIFx0YWIgNTUg N1x0YWIgNzEgR1x0YWIgODcgV1x0YWIgMTAzIGdcdGFiIDExOSB3DQpccGFyIDQwIChcdGFi IDU2IDhcdGFiIDcyIEhcdGFiIDg4IFhcdGFiIDEwNCBoXHRhYiAxMjAgeA0KXHBhciA0MSAp XHRhYiA1NyA5XHRhYiA3MyBJXHRhYiA4OSBZXHRhYiAxMDUgaVx0YWIgMTIxIHkNClxwYXIg NDIgKlx0YWIgNTggOlx0YWIgNzQgSlx0YWIgOTAgWlx0YWIgMTA2IGpcdGFiIDEyMiB6DQpc cGFyIDQzICtcdGFiIDU5IDtcdGFiIDc1IEtcdGFiIDkxIFtcdGFiIDEwNyBrXHRhYiAxMjMg XHsNClxwYXIgNDQgLFx0YWIgNjAgPFx0YWIgNzYgTFx0YWIgOTIgXFxcdGFiIDEwOCBsXHRh YiAxMjQgfA0KXHBhciA0NSBcZW5kYXNoIFx0YWIgNjEgPVx0YWIgNzcgTVx0YWIgOTMgXVx0 YWIgMTA5IG1cdGFiIDEyNSBcfQ0KXHBhciA0NiAuXHRhYiA2MiA+XHRhYiA3OCBOXHRhYiA5 NCBeXHRhYiAxMTAgblx0YWIgMTI2IH4NClxwYXIgNDcgL1x0YWIgNjMgP1x0YWIgNzkgT1x0 YWIgOTUgX1x0YWIgMTExIG8gXHRhYiAxMjcgZGVsZXRlIA0KXHBhciANClxwYXIgVGhpcyBp cyB0aGUgb3JpZ2luYWwgYmFzaWMgc2V0IG9mIGNoYXJhY3RlcnMgaW4gdGhlIFVTIEFTQ0lJ IFN0YW5kYXJkIGFuZCB3aGljaCBpcyBub3cgZW5jYXBzdWxhdGVkIGluIHRoZSBmaXJzdCBo YWxmIG9mIHdoYXQgaXMgY2FsbGVkIHRoZSBJU08gODg1OSBTdGFuZGFyZC4NClxwYXIgDQpc cGFyIE5vdGUgdGhhdCB0aGlzIHNldCBvZiBjaGFyYWN0ZXJzIGRvZXMgbm90IGluY2x1ZGUg dGhlIHBvdW5kIHNpZ24sIFwnYTMsIG9yIHRoZSBFdXJvIHNpZ24gKHNlZSBiZWxvdykgYW5k IG9idmlvdXNseSBkb2VzIG5vdCBpbmNsdWRlIGFueSBhY2NlbnRlZCBjaGFyYWN0ZXJzIHVz ZWQgaW4gb3RoZXIgY291bnRyaWVzIG9yIGFueSBzcGVjaWFsIGNoYXJhY3RlcnMgbGlrZSB0 aGUgc3ltYm9scyBmb3IgZGl2aXNpb24sIFwnZjcNCiwgYW5kIG11bHRpcGxpY2F0aW9uLCBc J2Q3LCBhbmQgb3RoZXIgdXNlZnVsIGNoYXJhY3RlcnMgd2hpY2ggd2Ugd291bGQgbGlrZSB0 byBiZSBhYmxlIHRvIHVzZSBpbiBvdXIgd3JpdGluZyBnZW5lcmFsbHkgYW5kLCBieSBleHRl bnNpb24sIHRvIGJlIGFibGUgdG8gc2VuZCBpbiBvdXIgZS1tYWlscy4NClxwYXIgDQpccGFy IFRoaXMgaGFzIGJlZW4gYWNoaWV2ZWQgZm9yIFdlc3Rlcm4gRXVyb3BlYW4gY291bnRyaWVz IChhbmQgb3RoZXJzKSBieSBhbGxvY2F0aW5nIGNoYXJhY3RlcnMgdG8gdGhvc2UgdW51c2Vk IHZhbHVlcyBpbiBhIGJ5dGUgZnJvbSAxNjAgdG8gMjU1IChWYWx1ZXMgZnJvbSAxMjggdG8g MTU5IHdlcmUgbm90IGFsbG9jYXRlZCBpbml0aWFsbHkgYXMgaXQgd2FzIHRob3VnaHQgdGhl eSBtaWdodCBjYXVzZSBhIGNsYXNoIHdpdGggdGhlIGNvcnJlc3Bvbg0KZGluZyBjb250cm9s IGNoYXJhY3RlcnMgaW4gdGhlIGJhc2ljIEFTQ0lJIGNoYXJhY3RlciBzZXQgZnJvbSAwIHRv IDMxLiBIb3dldmVyLCAxMjggaGFzIG5vdyBiZWVuIGFsbG9jYXRlZCB0byB0aGUgRXVybyBz aWduLCBcJzgwKS4gVGhlc2UgYXJlIGVuY2Fwc3VsYXRlZCBpbiB0aGUgSVNPIDg4NTktMSAo TGF0aW4gMSkgc3RhbmRhcmQgY2hhcmFjdGVyIHNldA0KIGluIHdoYXQgaXMgc29tZXRpbWVz IHJlZmVycmVkIHRvIGFzIHRoZSB1cHBlciBoYWxmIG9mIHRoZSBzZXQgYXMgc2hvd24gYmVs b3cuIA0KXHBhciANClxwYXIgTm90ZSB0aGF0IHRoZSBFdXJvIHNpZ24gaGFzIG5vdyBiZWVu IGluY2x1ZGVkIG9mZmljaWFsbHkgaW4gSVNPIDg4NTktMTUsIGFuIGFsdGVybmF0aXZlIFdl c3Rlcm4gRXVyb3BlYW4gY2hhcmFjdGVyIHNldCwgd2l0aCB0aGUgdmFsdWUgb2YgMTY0LiBU aGlzIGlzIHdoYXQgd2UgZ2V0IGZyb20gdGhlIFVLIGFuZCBVUyBrZXlib2FyZHMuIFRoaXMg cHJvZHVjZXMgdGhlIGludGVybmF0aW9uYWwgY3VycmVuY3kgc2lnbiwgXCdhNA0KLCBpbiBJ U08gODg1OS0xLCBwcm92aWRlIE1JTUUgZW5jb2Rpbmcgd2l0aCBRdW90ZWQgUHJpbnRhYmxl IGhhcyBiZWVuIHNldCwgb3RoZXJ3aXNlIGl0IHJldmVydHMgdG8gdGhlIGRvbGxhciBzaWdu ISkNClxwYXIgDQpccGFyIA0KXHBhciAxNjAgIFx0YWIgMTc2IFwnYjBcdGFiIDE5MiBcJ2Mw XHRhYiAyMDggXCdkMFx0YWIgMjI0IFwnZTBcdGFiIDI0MCBcJ2YwIA0KXHBhciAxNjEgXCdh MVx0YWIgMTc3IFwnYjFcdGFiIDE5MyBcJ2MxXHRhYiAyMDkgXCdkMVx0YWIgMjI1IFwnZTFc dGFiIDI0MSBcJ2YxDQpccGFyIDE2MiBcJ2EyXHRhYiAxNzggXCdiMlx0YWIgMTk0IFwnYzJc dGFiIDIxMCBcJ2QyXHRhYiAyMjYgXCdlMlx0YWIgMjQyIFwnZjINClxwYXIgMTYzIFwnYTNc dGFiIDE3OSBcJ2IzXHRhYiAxOTUgXCdjM1x0YWIgMjExIFwnZDNcdGFiIDIyNyBcJ2UzXHRh YiAyNDMgXCdmMw0KXHBhciAxNjQgXCdhNFx0YWIgMTgwIFwnYjRcdGFiIDE5NiBcJ2M0XHRh YiAyMTIgXCdkNFx0YWIgMjI4IFwnZTRcdGFiIDI0NCBcJ2Y0DQpccGFyIDE2NSBcJ2E1XHRh YiAxODEgXCdiNVx0YWIgMTk3IFwnYzVcdGFiIDIxMyBcJ2Q1XHRhYiAyMjkgXCdlNVx0YWIg MjQ1IFwnZjUNClxwYXIgMTY2IFwnYTZcdGFiIDE4MiBcJ2I2XHRhYiAxOTggXCdjNlx0YWIg MjE0IFwnZDZcdGFiIDIzMCBcJ2U2XHRhYiAyNDYgXCdmNg0KXHBhciAxNjcgXCdhN1x0YWIg MTgzIFwnYjdcdGFiIDE5OSBcJ2M3XHRhYiAyMTUgXCdkN1x0YWIgMjMxIFwnZTdcdGFiIDI0 NyBcJ2Y3IA0KXHBhciAxNjggXCdhOFx0YWIgMTg0IFwnYjhcdGFiIDIwMCBcJ2M4XHRhYiAy MTYgXCdkOFx0YWIgMjMyIFwnZThcdGFiIDI0OCBcJ2Y4DQpccGFyIDE2OSBcJ2E5XHRhYiAx ODUgXCdiOVx0YWIgMjAxIFwnYzlcdGFiIDIxNyBcJ2Q5XHRhYiAyMzMgXCdlOVx0YWIgMjQ5 IFwnZjkNClxwYXIgMTcwIFwnYWFcdGFiIDE4NiBcJ2JhXHRhYiAyMDIgXCdjYVx0YWIgMjE4 IFwnZGFcdGFiIDIzNCBcJ2VhXHRhYiAyNTAgXCdmYQ0KXHBhciAxNzEgXCdhYlx0YWIgMTg3 IFwnYmJcdGFiIDIwMyBcJ2NiXHRhYiAyMTkgXCdkYlx0YWIgMjM1IFwnZWJcdGFiIDI1MSBc J2ZiDQpccGFyIDE3MiBcJ2FjXHRhYiAxODggXCdiY1x0YWIgMjA0IFwnY2NcdGFiIDIyMCBc J2RjXHRhYiAyMzYgXCdlY1x0YWIgMjUyIFwnZmMNClxwYXIgMTczIFwnYWRcdGFiIDE4OSBc J2JkXHRhYiAyMDUgXCdjZFx0YWIgMjIxIFwnZGRcdGFiIDIzNyBcJ2VkXHRhYiAyNTMgXCdm ZA0KXHBhciAxNzQgXCdhZVx0YWIgMTkwIFwnYmVcdGFiIDIwNiBcJ2NlXHRhYiAyMjIgXCdk ZVx0YWIgMjM4IFwnZWVcdGFiIDI1NCBcJ2ZlDQpccGFyIDE3NSBcJ2FmXHRhYiAxOTEgXCdi Zlx0YWIgMjA3IFwnY2ZcdGFiIDIyMyBcJ2RmXHRhYiAyMzkgXCdlZlx0YWIgMjU1IFwnZmYN ClxwYXIgDQpccGFyIEhvd2V2ZXIsIHJlbWVtYmVyaW5nIHRoYXQgb25seSB2YWx1ZXMgaW4g dGhlIHJhbmdlIDAgdG8gMTI3IGNhbiBiZSBzZW50IHJlbGlhYmx5IHRocm91Z2ggdGhlIGlu dGVybmV0LCB0aGUgcHJvYmxlbSBpcyBob3cgY291bGQgd2Ugc2VuZCB0aGVzZSBjaGFyYWN0 ZXJzIGluIG91ciBlLW1haWxzPw0KXHBhciANClxwYXIgDQpccGFyIH17XGJcZjExMSAzLiBN SU1FIGVuY29kaW5nDQpccGFyIH17XGYxMTFcZnMyMCANClxwYXIgVGhlIHNvbHV0aW9uIGlz IHRvIHVzZSBNSU1FIGVuY29kaW5nIHdoaWNoIHdhcyBkZXZpc2VkIGJ5IHRoZSBJbnRlcm5l dCBDb25zb3J0aXVtIHRvIG92ZXJjb21lIHRoZSBVUyBBU0NJSSBib3R0bGVuZWNrLiBUaGVy ZSBhcmUgdHdvIHNjaGVtZXMgd2l0aGluIE1JTUUgZm9yIHRyYW5zbWl0dGluZyBlLW1haWxz IGNvbnRhaW5pbmcgY2hhcmFjdGVycyB3aXRoIHZhbHVlcyBmcm9tIDEyOCB0byAyNTUuIE9u ZSBpcyBjYWxsZWQgQmFzZTY0IChub3QgDQpyZWNvbW1lbmRlZCBoZXJlKSB3aGljaCBlbnRh aWxzIGNvbnZlcnRpbmcgZXZlcnkgdGhyZWUgOCBiaXQgY2hhcmFjdGVycyBpbnRvIGZvdXIg NyBiaXQgY2hhcmFjdGVycyB1c2luZyBvbmx5IHRoZSBiYXNpYyBBU0NJSSBjaGFyYWN0ZXJz ICh2YWx1ZXMgMCBcZW5kYXNoIA0KIDEyNyksIHRoZSBvdGhlciBpcyBjYWxsZWQgUXVvdGVk IFByaW50YWJsZSB3aGljaCBlbnRhaWxzIGNvbnZlcnRpbmcgb25seSB0aG9zZSBjaGFyYWN0 ZXJzIHdpdGggYSB2YWx1ZSBvdmVyIDEyNyBpbnRvIGEgdHJpcGxlIG9mIGNoYXJhY3RlcnMg Y29uc2lzdGluZyBvZiBhbiBlcXVhbHMgc2lnbiBmb2xsb3dlZCANCmJ5IHRoZSB0d28gY2hh cmFjdGVyIGhleGFkZWNpbWFsIChzZWUgZ2xvc3NhcnkgYmVsb3cpIHZhbHVlIG9mIHRoZSBk ZWNpbWFsIHZhbHVlIG9mIHRoYXQgY2hhcmFjdGVyIChhbGwgdGhlc2UgcG9zc2libGUgY2hh cmFjdGVycyBhcmUgaW4gdGhlIGJhc2ljIEFTQ0lJIGNoYXJhY3RlciBzZXQgd2l0aCB2YWx1 ZXMgMCBcZW5kYXNoICAxMjcpLiBGb3IgZXhhbXBsZSwgXCdhMw0KNTAuMDAgd291bGQgYmUg Y29kZWQgYXMgPUEzNTAuMDAgd2l0aCB0aGUgPUEzIHJlcHJlc2VudGluZyB0aGUgcG91bmQg c2lnbiB3aXRoIGEgdmFsdWUgb2YgQTMgPSAxNjMgKHNlZSBhYm92ZSkuIFRoaXMgd291bGQg YmUgY29udmVydGVkIGF1dG9tYXRpY2FsbHkgYmFjayB0byBhIHBvdW5kIHNpZ24gYnkgdGhl IHJlY2lwaWVudFxycXVvdGUgDQpzIGUtbWFpbCBzeXN0ZW0gUFJPVklERUQgdGhhdCB0aGUg U0VOREVSIGhhZCBzcGVjaWZpZWQgTUlNRSBlbmNvZGluZyB3aXRoIFF1b3RlZCBQcmludGFi bGUgc2V0IGFzIGRlc2NyaWJlZCBhYm92ZS4NClxwYXIgDQpccGFyIEEgZGlzYWR2YW50YWdl IG9mIEJhc2U2NCBpcyB0aGF0IHRoZSBsZW5ndGggb2YgdGhlIG1lc3NhZ2UgaXMgaW5jcmVh c2VkIGNvbnNpZGVyYWJseSB3aGV0aGVyIHRoZXJlIGFyZSBhbnkgYWNjZW50ZWQgb3Igc3Bl Y2lhbCBjaGFyYWN0ZXJzIGluIHRoZSBtZXNzYWdlIG9yIG5vdCwgd2hlcmVhcyB0aGUgYWR2 YW50YWdlIG9mIFF1b3RlZCBQcmludGFibGUgaXMgdGhhdCBvbmx5IHRob3NlIGFjY2VudGVk IG9yIHNwZWNpYWwgY2hhcmFjdGVycyB3aQ0KdGggdmFsdWVzIG92ZXIgMTI3IGFyZSBjb252 ZXJ0ZWQgZnJvbSBvbmUgY2hhcmFjdGVyIHRvIHRocmVlIGNoYXJhY3RlcnMuDQpccGFyIA0K XHBhciBUaGUgbWFqb3JpdHkgb2YgY3VycmVudCBlLW1haWwgc3lzdGVtcyBoYXZlIHRoZSBm YWNpbGl0eSB0byBzZXQgdGhlaXIgc3lzdGVtIHRvIHNlbmQgZS1tYWlscyB1c2luZyBNSU1F IGVuY29kaW5nIHdpdGggUXVvdGVkIFByaW50YWJsZSBzZXQuIFRoZSBwb2ludCB0byByZW1l bWJlciBhYm91dCB0aGlzIGlzIHRoYXQgb25jZSBzZXQgdGhlcmUgaXMgbm8gbmVlZCB0byBj aGFuZ2UgdGhlIHNldHRpbmcgZnJvbSBvbmUgbWVzc2FnZSB0byB0aGUgbmUNCnh0LiBJZiB0 aGVyZSBhcmUgbm8gYWNjZW50ZWQgb3Igc3BlY2lhbCBjaGFyYWN0ZXJzIGluIGEgbWVzc2Fn ZSwgTUlNRSBlbmNvZGluZyB3aXRoIFF1b3RlZCBQcmludGFibGUgd2lsbCBub3QgYmUgaW52 b2tlZC4gSW5kZWVkLCB0aGVyZSBpcyBhbiBhcmd1bWVudCB0aGF0IHRoaXMgc2V0dGluZyBz aG91bGQgYmUgdGhlIGRlZmF1bHQgc2V0dGluZyBmb3IgYWxsIHRleHQgYmFzZWQgZS1tYWls IHN5c3RlbXMuDQpccGFyIA0KXHBhciBIb3cgdG8gc2V0IHNvbWUgZS1tYWlsIHN5c3RlbXMg Zm9yIE1JTUUgZW5jb2Rpbmcgd2l0aCBRdW90ZWQgUHJpbnRhYmxlIHNldCBpcyBkZXNjcmli ZWQgdW5kZXIgXGxkYmxxdW90ZSBTZXR0aW5ncyBmb3Igc29tZSBFLU1haWwgU3lzdGVtc1xy ZGJscXVvdGUgIGFib3ZlLg0KXHBhciANClxwYXIgDQpccGFyIA0KXHBhciANClxwYXIgDQpc cGFyIA0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGJcZjExMSA0LiBPdGhlciBDaGFyYWN0 ZXIgU2V0cw0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFcZnMyMCANClxwYXIgfVxw YXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpUaGUgYWJvdmUgZGlzY3Vzc2lvbiBoYXMg YXNzdW1lZCB0aGF0IHRoZSBzdGFuZGFyZCBJU08gODg1OS0xIChMYXRpbjEgZm9yIFdlc3Rl cm4gRXVyb3BlKSBpcyBiZWluZyB1c2VkLiBJU08gODg1OS0xNSBpcyBhIGxhdGVyIG1vZGlm aWNhdGlvbiBvZiBJU08gODg1OS0xIChMYXRpbjkgZm9yIFdlc3Rlcm4gRXVyb3BlKSBpbiB3 aGljaCB0aGUgRXVybyBoYXMgYmVlbiBpbmNvcnBvcmF0ZWQgYW5kIHNvbWUgaGl0aGVydG8g bWlzc2luZyBzcGVjaWENCmwgY2hhcmFjdGVycyBmb3IgRnJlbmNoLCBhdCB0aGUgZXhwZW5z ZSBvZiA3IHByZXZpb3VzbHkgbGVzcyB1c2VkIGNoYXJhY3RlcnMuDQpccGFyIH1ccGFyZCBc d2lkY3RscGFyIHtcZjExMVxmczIwIA0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFc ZnMyMCANCkhvd2V2ZXIsIHRoZXJlIGFyZSBvdGhlciBjaGFyYWN0ZXIgc2V0cyB3aXRoaW4g dGhlIElTTyA4ODU5IHJhbmdlLCB3aGljaCB5b3VyIGUtbWFpbCBzeXN0ZW0gbWF5IGJlIGFi bGUgdG8gdXNlIGZvciBvdGhlciBsYW5ndWFnZXMsIHN1Y2ggYXMgQ3lyaWxsaWMgKFJ1c3Np YW4gZXRjLikgb3IgR3JlZWssIHdpdGggSVNPIHN0YW5kYXJkIG51bWJlcnMsIHN1Y2ggYXMg SVNPIDg4NTktNSwgSVNPIDg4NTktNyBldGMuIChhbmQgYWxzbyBVTklDT0RFIA0KKFVURi03 IG9yIFVURi04KSBub3QgZGVzY3JpYmVkIGhlcmUpLg0KXHBhciB9XHBhcmQgXHdpZGN0bHBh ciB7XGYxMTFcZnMyMCANClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgSW4g cHJhY3RpY2UsIGluIHRoZSBVU0EgYW5kIG1vc3Qgb2YgRXVyb3BlIHRoZSBJU08gODg1OSBy YW5nZSBpcyBub3JtYWxseSB1c2VkICgyMDAyKSBmb3IgZS1tYWlscyBidXQgVU5JQ09ERSBt YXkgYmUgdXNlZCBmb3IgbWFueSBFYXN0ZXJuIGxhbmd1YWdlcyB3aG9zZSBjaGFyYWN0ZXJz IGNhbm5vdCBiZSBhY2NvbW1vZGF0ZWQgd2l0aGluIGEgbGltaXQgb2YgMjU2IGNoYXJhY3Rl cnMuDQoNClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpccGFyIH1ccGFy ZCBcd2lkY3RscGFyIHtcZjExMVxmczIwIFRoZSBzdGFuZGFyZCBJU08gODg1OSByYW5nZSBh bGwgdXNlIHRoZSBzYW1lIGJhc2ljIFVTIEFTQ0lJIGNoYXJhY3RlciBzZXQgd2l0aCB2YWx1 ZXMgZnJvbSAwIFxlbmRhc2ggDQogMTI3IGJ1dCB0aGV5IGVhY2ggaGF2ZSBhIGZ1cnRoZXIg c2V0IG9mIGNoYXJhY3RlcnMgcGVjdWxpYXIgdG8gdGhlaXIgbGFuZ3VhZ2VzIGluIHRoZSB1 cHBlciBoYWxmIG9mIHRoZSBzZXQsIHdpdGggdGhlIHZhbHVlcyAxNjAgXGVuZGFzaCAgMjU1 LiBZb3UgY2FuIHNlZSBhIGxpc3Qgb2YgdGhlc2UgYXQgaHR0cDovL2N6eWJvcnJhLmNvbS9j aGFyc2V0cy9pc284ODU5Lmh0bWwNClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZz MjAgDQpccGFyIH1ccGFyZCBcd2lkY3RscGFyIHtcZjExMVxmczIwIFRvIHNlbmQgYW4gZS1t YWlsIGNvbnRhaW5pbmcgdGhlc2UgY2hhcmFjdGVycyB5b3UgbXVzdCwgb2YgY291cnNlLCBj b250aW51ZSB0byB1c2UgTUlNRSBlbmNvZGluZyB3aXRoIFF1b3RlZCBQcmludGFibGUgc2V0 LCBidXQgaW4gYWRkaXRpb24geW91IHNob3VsZCBzcGVjaWZ5IHRoZSBpbnRlDQpuZGVkIG1l YW5pbmcgb2YgdGhlIHZhbHVlcyBiZWluZyBzZW50IGJ5IHNwZWNpZnlpbmcgdGhlIGNoYXJh Y3RlciBzZXQgaW50ZW5kZWQgYXMgSVNPIDg4NTktMiwgb3IgSVNPIDg4NTktMyBldGMuIGFz IHJlcXVpcmVkLg0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFcZnMyMCANClxwYXIg fVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgSG93IHRvIHNldCBzb21lIGUtbWFpbCBz eXN0ZW1zIGZvciBkaWZmZXJlbnQgY2hhcmFjdGVyIHNldHMgaXMgZGVzY3JpYmVkIHVuZGVy IFxsZGJscXVvdGUgU2V0dGluZ3MgZm9yIHNvbWUgZS1tYWlsIHN5c3RlbXNccmRibHF1b3Rl ICBhYm92ZS4NClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpccGFyIH1c cGFyZCBcd2lkY3RscGFyIHtcYlxmMTExIDUuIENoYXJhY3RlciBlbnRyeQ0KXHBhciB9XHBh cmQgXHdpZGN0bHBhciB7XGYxMTFcZnMyMCANClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xm MTExXGZzMjAgRGlmZmVyZW50IGNvdW50cmllcyBoYXZlIGRpZmZlcmVudCBrZXlib2FyZHMg ZGVzaWduZWQgZm9yIHRoZWlyIGxhbmd1YWdlcyBhbmQgdGhlc2UgZW5hYmxlIHVzZXJzIHRv IGVudGVyIHRoZWlyIGxhbmd1YWdlXHJxdW90ZSANCnMgYWNjZW50ZWQgYW5kIHNwZWNpYWwg Y2hhcmFjdGVycyBieSB1c2luZyBhcHByb3ByaWF0ZWx5IG1hcmtlZCBrZXlzLiBUaGUga2V5 Ym9hcmQsIGNvbWJpbmVkIHdpdGggYXBwcm9wcmlhdGUgbGFuZ3VhZ2UgdGFiDQpsZXMgYWxy ZWFkeSBzZXQgdXAgaW4gdGhlIGNvbXB1dGVyIHN5c3RlbSwgZ2VuZXJhdGVzIHRoZSBjb3Jy ZWN0IGNvbXB1dGVyIHZhbHVlcyBmb3IgdGhvc2UgY2hhcmFjdGVycyBhYm92ZSB0aGUgdmFs dWUgb2YgMTI3LiBGb3IgaW5zdGFuY2UsIHRoZSBVSyBrZXlib2FyZCBoYXMga2V5cyBmb3Ig dGhlIHBvdW5kIHNpZ24gKDE2MykgYW5kIHRoZSBFdXJvICgxNjQpLCB0aGUgRnJlbmNoIGtl eWJvYXJkIGhhcyBrZXlzIGZvciBcbGRibHF1b3RlIGENClxyZGJscXVvdGUgIGFjdXRlICgx OTMgYW5kIDIyNSkgYW5kIFxsZGJscXVvdGUgY1xyZGJscXVvdGUgIGNlZGlsbGEgKDE5OSBh bmQgMjMxKSBldGMuDQpccGFyIH1ccGFyZCBcd2lkY3RscGFyIHtcZjExMVxmczIwIA0KXHBh ciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFcZnMyMCBIb3dldmVyLCB0aGUgcXVlc3Rpb24g YXJpc2VzIGFib3V0IHdoYXQgdG8gZG8gaWYgeW91IHdhbnQgdG8gc2VuZCBhbiBhY2NlbnRl ZCBvciBzcGVjaWFsIGNoYXJhY3RlciB3aGljaCBpcyBub3QgcmVwcmVzZW50ZWQgb24geW91 ciBrZXlib2FyZC4gRm9yIGluc3RhbmNlLCBhIFVLIHVzZXIgbWF5IHdpc2ggdG8gc2VuZCBh biBlLW1haWwgdG8gYSBGcmVuY2ggY29sbA0KZWFndWUgaW4gRnJlbmNoIHJlcXVpcmluZyB0 aGUgdXNlIG9mIHNvbWUgYWNjZW50ZWQgY2hhcmFjdGVycyBub3Qgc2hvd24gb24gYSBVSyBr ZXlib2FyZC4NClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpccGFyIH1c cGFyZCBcd2lkY3RscGFyIHtcZjExMVxmczIwIA0KU29tZSB0ZXh0IGVkaXRpbmcgc3lzdGVt cyBkbyBoYXZlIGZhY2lsaXRpZXMgZm9yIGRpc3BsYXlpbmcgdGhlIGV4dHJhIGNoYXJhY3Rl cnMgYXZhaWxhYmxlIGZvciBzZWxlY3Rpb24gKGFuZCB0aGlzIHdvdWxkIGJlIGEgdXNlZnVs IGFkZGl0aW9uIHRvIGFuIGUtbWFpbCBzeXN0ZW0pIGJ1dCB3ZSBhcmUgY29uc2lkZXJpbmcg Y29tcG9zaW5nIG1lc3NhZ2VzIGluIGEgc3RhbmRhcmQgd2F5IGluIGEgdmFyaWV0eSBvZiBl LW1haWwgc3lzdGVtcy4gDQpUaGUgc2ltcGxlc3Qgc3RhbmRhcmQgd2F5IG9mIGVudGVyaW5n IGNoYXJhY3RlcnMgbm90IGFwcGVhcmluZyBvbiB5b3VyIGtleWJvYXJkIGlzIHRvIHVzZSBh IHN0YW5kYXJkIGtleWJvYXJkIGZhY2lsaXR5IGFzIGZvbGxvd3MuDQpccGFyIH1ccGFyZCBc d2lkY3RscGFyIHtcZjExMVxmczIwIA0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFc ZnMyMCBIb2xkIGRvd24gdGhlIEFsdCBrZXkgYW5kLCB1c2luZyB0DQpoZSBrZXlwYWQsIGVu dGVyIHRoZSBkaWdpdCB6ZXJvIGZvbGxvd2VkIGJ5IHRoZSB0aHJlZSBkaWdpdCB2YWx1ZSBv ZiB0aGUgY2hhcmFjdGVyIHJlcXVpcmVkLiBOb3RoaW5nIGhhcHBlbnMgYXQgZmlyc3QgYnV0 IHdoZW4geW91IHRha2UgeW91ciBmaW5nZXIgb2ZmIHRoZSBBbHQga2V5IHRoZSByZXF1aXJl ZCBjaGFyYWN0ZXIgYXBwZWFycyBvbiB0aGUgc2NyZWVuLiBGb3IgZXhhbXBsZSwgXCdmNA0K IGlzIEFsdC8wMjQ0IC4gQmV3YXJlIHRoYXQgZm9yIHRoaXMgdG8gd29yayB3aXRoIHNvbWUg ZS1tYWlsIHN5c3RlbXMgKGUuZy4gT3V0bG9vayBFeHByZXNzIGFuZCBFdWRvcmEpIHRoZSBO dW1Mb2NrIGxpZ2h0IG5lZWRzIHRvIGJlIG9uLg0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7 XGYxMTFcZnMyMCANClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgT2YgY291 cnNlLCB5b3UgbmVlZCB0byBrbm93IHRoZSBkZWNpbWFsIHZhbHVlIG9mIHRoZSBjaGFyYWN0 ZXIgcmVxdWlyZWQuIFRoZXJlIGFyZSBhIG51bWJlciBvZiB3ZWJzaXRlcyB3aGljaCBjb250 YWluIGxpc3RzIG9mIHRoZSBjb250ZW50cyBvZiBtb3N0LCBpZiBub3QgYWxsLCBvZiB0aGUg SVNPIDg4NTktbiByYW5nZSBvZiBjaGFyYWN0ZXJzIHNldHMgKHN1Y2ggYXMgQ3p5Ym9ycmEN ClxycXVvdGUgcyBzaXRlIG1lbnRpb25lZCBhYm92ZSkuIFRoZSBjdXJyZW50IGhpZ2hlc3Qg dmFsdWUgb2YgbiBpcyAxNSAoMjAwMiksIGJ1dCBuZXcgc2V0cyBhcmUgYWRkZWQgZnJvbSB0 aW1lIHRvIHRpbWUuIFRoZSBiZXN0IHRoaW5nIHRvIGRvIGlzIHRvIHdyaXRlIG91dCB0aGUg c2V0IHlvdSBhcmUgZ29pbmcgdG8gdXNlIGZyZXF1DQplbnRseSBpbiBtdWNoIHRoZSBzYW1l IGZvcm1hdCBhcyBtaW5lIGFib3ZlIGFuZCBzYXZlIGl0IGluIGEgZmlsZSB3aGljaCB5b3Ug Y2FuIHRoZW4gaGF2ZSBvcGVuIGF0IHRoZSBzYW1lIHRpbWUgYXMgY29tcG9zaW5nIHlvdXIg ZS1tYWlscyBzbyB5b3UgY2FuIHJlZmVyIHRvIGl0IGVhc2lseS4gSWYgeW91IHNlbmQgYSBs b3Qgb2YgdGhlc2UgZS1tYWlscyBJXHJxdW90ZSANCm0gc3VyZSB5b3Ugd2lsbCBiZWdpbiB0 byByZW1lbWJlciB0aGUgdmFsdWVzIG9mIHRoZSBtb3N0IGZyZXF1ZW50bHkgdXNlZCBjaGFy YWN0ZXJzIGFueXdheS4NClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpc cGFyIH1ccGFyZCBcd2lkY3RscGFyIHtcYlxmMTExIDYuIFJlY2VpdmluZyBlLW1haWxzDQpc cGFyIH1ccGFyZCBcd2lkY3RscGFyIHtcZjExMVxmczIwIA0KXHBhciB9XHBhcmQgXHdpZGN0 bHBhciB7XGYxMTFcZnMyMCANClNvbWV0aW1lcyB5b3Ugd2lsbCByZWNlaXZlIGUtbWFpbHMg ZnJvbSBhYnJvYWQgd2l0aCBzdHJhbmdlIGNoYXJhY3RlcnMgd2hlcmUgYWNjZW50ZWQgb3Ig c3BlY2lhbCBjaGFyYWN0ZXJzIHNob3VsZCBiZS4gV2l0aCB0aGUgYWJvdmUga25vd2xlZGdl IHlvdSB3aWxsIGJlIGFibGUgdG8gZGlhZ25vc2Ugd2hhdCBoYXMgZ29uZSB3cm9uZyB3aGlj aCB3b3VsZCBlbmFibGUgeW91IHRvIGNvcnJlY3QgdGhlIHRleHQgaWYgbmVjZXNzYXJ5IGFu ZC9vcg0KIGFkdmlzZSB0aGUgc2VuZGVyIG9mIHdoYXQgaGFzIGhhcHBlbmVkIGFuZCB3aGF0 IHRvIGRvIGFib3V0IGl0Lg0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFcZnMyMCAN ClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgRm9yIGluc3RhbmNlIGlmIHlv dSByZWNlaXZlIHRoZSB3b3JkIGZyYW5nYWlzIGZyb20gYSBGcmVuY2ggY29ycmVzcG9uZGVu dCBpdCBsb29rcyBhcyB0aG91Z2ggdGhlIGxldHRlciBnICgxMDMpIHNob3VsZCBoYXZlIGJl ZQ0KbiBhIGMgY2VkaWxsYSAoMjMxKSwgdGhlIGRpZmZlcmVuY2UgYmV0d2VlbiB0aGVpciB2 YWx1ZXMgaW4gdGhlIElTTyA4ODU5LTEgdGFibGUgYmVpbmcgMTI4LiBUaGlzIHdvdWxkIGlu ZGljYXRlIHRoYXQgdGhlIGUtbWFpbCBoYWQgYmVlbiBzZW50IFxsZGJscXVvdGUgYXMgaXNc cmRibHF1b3RlIA0KIHdpdGhvdXQgTUlNRSBlbmNvZGluZyBhbmQgUXVvdGVkIFByaW50YWJs ZSBzZXQgKG9yIGF0IGxlYXN0IHdpdGhvdXQgUXVvdGVkIFByaW50YWJsZSBzZXQpIHNvIHRo YXQgdGhlIGhpZ2ggb3JkZXIgYml0IG9mIHRoZSA4IGJpdCBieXRlIGhhZCBiZWVuIGxvc3Qg ZHVyaW5nIHRyYW5zbWlzc2lvbiB0aHJvdWdoIHRoZSBpbnRlcm5ldC4NClxwYXIgfVxwYXJk IFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpccGFyIH1ccGFyZCBcd2lkY3RscGFyIHtcZjEx MVxmczIwIElmIHlvdSBleGFtaW5lIHRoZSBmaXJzdCwgb3IgbG93ZXIgaGFsZiBvZiBhbiBJ U08gODg1OSBjaGFyYWN0ZXIgc2V0IChhbHdheXMgdGhlIHNhbWUgYXMgVVMgQVNDSUkpIGFu ZCBjb21wYXJlIGl0IHdpdGggdGhlIHNlY29uZCwgb3IgdXBwZXIgaGFsZiBvZiB0aGUgY2hh cmFjdGVyIHNldCwgeW91DQogY2FuIHNlZSB0aGUgY29ycmVzcG9uZGVuY2UgYmV0d2VlbiB0 aGUgY2hhcmFjdGVycyB3aGljaCBkaWZmZXIgaW4gdmFsdWUgYnkgMTI4IChmb3IgdGhlIHRl Y2hub3BoaWxlcywgdGhlIHZhbHVlIG9mIHRoZSBoaWdoIG9yZGVyIGJpdCB3aGljaCBjb3Vs ZCBiZSBsb3N0IGluIHRyYW5zbWlzc2lvbiB0aHJvdWdoIHRoZSBpbnRlcm5ldCkuDQpccGFy IH1ccGFyZCBcd2lkY3RscGFyIHtcZjExMVxmczIwIA0KXHBhciB9XHBhcmQgXHdpZGN0bHBh ciB7XGYxMTFcZnMyMCANClRoZSBzb2x1dGlvbiBpcyB0byBwZXJzdWFkZSB5b3VyIGNvcnJl c3BvbmRlbnQgdG8gc2V0IHRoZWlyIGUtbWFpbCBzeXN0ZW0gZm9yIE1JTUUgZW5jb2Rpbmcg d2l0aCBRdW90ZWQgUHJpbnRhYmxlIHNldCwgYW5kIHNwZWNpZnlpbmcgdGhlIHJlcXVpcmVk IElTTyA4ODU5LW4gY2hhcmFjdGVyIHNldC4gQWN0dWFsbHkgSVNPIDg4NTktMSBpcyBub3Jt YWxseSBhc3N1bWVkIGlmIG5vdCBzcGVjaWZpZWQsIGJ1dCBpdCBpcyBwcm9iYWJseSBzYWZl cg0KIHRvIGJlIHNwZWNpZmljLg0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFcZnMy MCANClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgT24gdGhlIG90aGVyIGhh bmQsIGl0IG1heSBiZSB0aGF0IHRoZSBzZW5kZXINCiBoYXMgc2VudCB5b3UgYW4gZS1tYWls IHVzaW5nIE1JTUUgZW5jb2Rpbmcgd2l0aCBRdW90ZWQgUHJpbnRhYmxlIHNldCBidXQgd2l0 aCBzb21lIG90aGVyIGNoYXJhY3RlciBzZXQgZnJvbSB0aGUgSVNPIDg4NTkgcmFuZ2UsIGZv ciBleGFtcGxlLCBJU08gODg1OS03IGZvciBHcmVlay4gT24gdGhlIGFzc3VtcHRpb24gdGhh dCB5b3VyIGUtbWFpbCBzeXN0ZW0gd2FzIHNldCB0byBleHBlY3QgSVNPIDg4NTktMSBieSBk ZWZhdWx0LCB5b3Ugd291bA0KZCBzZWUgRXVyb3BlYW4gYWNjZW50ZWQgY2hhcmFjdGVycyBk b3R0ZWQgYWJvdXQgd2hhdCBob3BlZnVsbHkgeW91IHdvdWxkIHJlY29nbmlzZSBhcyBHcmVl ayB3b3Jkcy4gSW4gdGhhdCBjYXNlIHlvdSB3b3VsZCByZWFsaXNlIHRoYXQgeW91IHdlcmUg YXQgbGVhc3Qgc2VlaW5nIHRoZSB1cHBlciBoYWxmIG9mIGFuIElTTyA4ODU5IGNoYXJhY3Rl ciBzZXQgYnV0IG5lZWRlZCB0byBzZXQgeW91ciBvd24gZS1tYWlsIHN5c3RlbSB0byB1c2Ug b25lDQogb2YgdGhlIG90aGVyIGNoYXJhY3RlciBzZXRzLCBpbiB0aGlzIGNhc2UgSVNPIDg4 NTktNyAoR3JlZWspLg0KXHBhciB9XHBhcmQgXHdpZGN0bHBhciB7XGYxMTFcZnMyMCANClxw YXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgVG8gY2hhbmdlIHlvdXIgY2hhcmFj dGVyIHNldCB0byBtYXRjaCB0aGUgc2VuZGVyXHJxdW90ZSBzIHlvdSB3b3VsZCBuZWVkIHRv IGNsaWNrIHNvbWV0aGluZyBsaWtlIFxsZGJscXVvdGUgVmlld1xyZGJscXVvdGUgIGFuZCBc bGRibHF1b3RlIENoYXJhY3RlciBTZXRccmRibHF1b3RlIA0KIHRvIHNlZSBhIGxpc3Qgb2Yg YXZhaWxhYmxlIGNoYXJhY3RlciBzZXRzIGFuZCB0aGVuIHRvIHNlbGVjdCB0aGUgb25lIHJl cXVpcmVkIHdoZW4geW91IHNob3VsZCBiZSBhYmxlIHRvIHNlZSB0aGUgdGV4dCBwcm9wZXJs eS4gSG93ZXZlciwgdA0KaGUgbnVtYmVyIG9mIGF2YWlsYWJsZSBjaGFyYWN0ZXIgc2V0cyBp biBhbnkgcGFydGljdWxhciBlLW1haWwgc3lzdGVtIGlzIHVzdWFsbHkgbGltaXRlZCBzbyB5 b3UgbWF5IGJlIHVubHVja3kuIEluIHdoaWNoIGNhc2UgeW91IG1pZ2h0IHdhbnQgdG8gbmVn b3RpYXRlIHdpdGggdGhlIHNlbmRlciB0byB1c2UgVVRGLTcgb3IgVVRGLTggKFVODQpJQ09E RSAtIG5vdCBkZXNjcmliZWQgaGVyZSkgaW5zdGVhZCwgb3IgY2hhbmdlIHRvIHVzZSBhbm90 aGVyIGUtbWFpbCBzeXN0ZW0gd2hpY2ggY2FuIHByb2Nlc3MgdGhlaXIgY2hhcmFjdGVyIHNl dCENClxwYXIgfVxwYXJkIFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpccGFyIH1ccGFyZCBc d2lkY3RscGFyIHtcZjExMVxmczIwIA0KSWYgdGhlIHNlbmRlciBoYWQgdXNlZCBVTklDT0RF IHRoZW4geW91IG1pZ2h0IHNlZSB0aGUgdGV4dCBzcHJlYWQgb3V0IHNvIHRoYXQgZXZlcnkg b3RoZXIgY2hhcmFjdGVyIG1hZGUgc2Vuc2UgYnV0IHRoZSBpbnRlcnZlbmluZyBjaGFyYWN0 ZXJzIHdlcmUgc3F1YXJlcyBvciBzb21lIG90aGVyIG1lYW5pbmdsZXNzIHN5bWJvbC4gSW4g dGhhdCBjYXNlIGNoYW5naW5nIHlvdXIgb3duIGNoYXJhY3RlciBzZXQgdG8gVVRGLTggb3Ig VVRGLTcgKFVODQpJQ09ERSkgbWlnaHQgY3VyZSB0aGUgcHJvYmxlbS4NClxwYXIgfVxwYXJk IFx3aWRjdGxwYXIge1xmMTExXGZzMjAgDQpccGFyIH1ccGFyZCBcd2lkY3RscGFyIHtcZjEx MVxmczIwIElmIHlvdSBhcmUgdW5hYmxlIHRvIHdvcmsgb3V0IHdoYXQgY2hhcmFjdGVyIHNl dCB0aGUgc2VuZGVyIHVzZWQgaXQgaXMgdXN1YWxseSBwb3NzaWJsZSB0byBleGFtaW5lIHRo ZSBzb3VyY2UgY29kZSBvZiBlLW1haWxzIGFzIGRlc2NyaWJlZCBiZWxvdyB0byBzZWUgd2hh dCBzZXR0aW5ncyB3ZXJlIGFjdHVhbGx5IHVzZWQuDQpccGFyIH1ccGFyZCBcd2lkY3RscGFy IHtcZjExMVxmczIwIA0KXHBhciBXaGVuZXZlciB5b3UgY2hhbmdlIHRoZSBzZXR0aW5nIG9m IHlvdXIgb3duIGNoYXJhY3RlciBzZXQgZG9uXHJxdW90ZSANCnQgZm9yZ2V0IHRvIHJlc3Rv cmUgeW91ciBkZWZhdWx0IGFmdGVyd2FyZHMgb3IgeW91IG1pZ2h0IG5vdCBiZSBhYmxlIHRv IHJlYWQgeW91ciBub3JtYWwgZS1tYWlscyBwcm9wZXJseSEgT3IgeW91IG1pZ2h0IGV2ZW4g aW5hZHZlcnRlbnRseSBsYXRlciBvbiBzZW5kIHlvdXIgb3duIGUtbWFpbHMgd2l0aCB0aGUg d3JvbmcgY29kaW5nLg0KXHBhciANClxwYXIgVGhlIGdvb2QgbmV3cyBpcyB0aGF0IHNvbWUg ZS1tYWlsIHN5c3RlbXMgd2lsbCBhdXRvbWF0aWNhbGx5IHNlbGVjdCB0aGUgY29ycmVjdCBj aGFyYWN0ZXIgc2V0IGZyb20gYW4gaW5jb21pbmcgbWVzc2FnZSwgYW5kIHByb3ZpZGVkIGl0 IGhhcyBiZWVuIHNlbnQgd2l0aCBNSU1FIGVuY29kaW5nIHdpdGggUXVvdA0KZWQgUHJpbnRh YmxlIHNldCwgaXQgc2hvdWxkIGRpc3BsYXkgY29ycmVjdGx5IHdpdGhvdXQgYW55IGludGVy dmVudGlvbiBvbiB5b3VyIHBhcnQuDQpccGFyIA0KXHBhciANClxwYXIgfXtcYlxmMTExIDcu IFNvdXJjZSBjb2RlIG9mIGUtbWFpbHMNClxwYXIgfXtcZjExMVxmczIwIA0KXHBhciBUaGUg b3JpZ2luYWwgc291cmNlIGNvZGUsIG9yIG1vcmUgaW1wb3J0YW50bHksIHRoZSB0cmFuc21p c3Npb24gaW5mb3JtYXRpb24gaW4gdGhlIGhlYWRpbmcgb2YgYW4gZS1tYWlsIG1lc3NhZ2Ug bm90IG5vcm1hbGx5IGRpc3BsYXllZCB3aXRoIHRoZSBtZXNzYWdlLCBjYW4gYmUgZXhhbWlu ZWQgaW4gbW9zdCBlLW1haWwgc3lzdGVtcy4gU2VlIGJlbG93IGZvciBob3cgdG8gZG8gdGhp cyBmb3Igc29tZSBvZiB0aGUgbW9yZSBwb3B1bGFyIGUtbWENCmlsIHN5c3RlbXMuDQpccGFy IA0KXHBhciBUaGUgaW5mb3JtYXRpb24gb2YgaW50ZXJlc3QgaGVyZSBpcyB3aGV0aGVyIE1J TUUgZW5jb2Rpbmcgd2FzIHNwZWNpZmllZCwgd2hldGhlciBRdW90ZWQgUHJpbnRhYmxlIHdh cyBzZXQsIGFuZCB0aGUgQ2hhcmFjdGVyIFNldCBzcGVjaWZpZWQsIGFuZCBhbGwgdXN1YWxs eSBhcHBlYXJpbmcgbmVhciB0aGUgYmVnaW5uaW5nIG9mIHRoZSBlLW1haWwsIG9yIHBvc3Np Ymx5IGluIGZyb250IG9mIHNlcGFyYXRlIHBhcnRzIG9mIHRoZSBlLW1haWwuDQoNClxwYXIg DQpccGFyIE1JTUUgZW5jb2RpbmcgdXN1YWxseSBzaG93biBhcyBcbGRibHF1b3RlIE1JTUUg VmVyc2lvbjogMS4wXHJkYmxxdW90ZSAsIG9yIHNpbWlsYXIuDQpccGFyIA0KXHBhciBRdW90 ZWQgUHJpbnRhYmxlIHNob3duIGJ5IFxsZGJscXVvdGUgWC1NWC1Db21tZW50OiBRVU9URUQt UFJJTlRBQkxFIG1lc3NhZ2UgYXV0b21hdGljYWxseSBkZWNvZGVkXHJkYmxxdW90ZSAsIG9y IHNpbWlsYXIuDQpccGFyIA0KXHBhciBDaGFyYWN0ZXIgc2V0IHNob3duIGJ5IFxsZGJscXVv dGUgQ29udGVudC1UeXBlOiB0ZXh0L3BsYWluOyBjaGFyc2V0PWlzby04ODU5LTFccmRibHF1 b3RlICwgb3Igc2ltaWxhci4NClxwYXIgDQpccGFyIFByb2NlZHVyZXMgZm9yIGRpc3BsYXlp bmcgdGhlIHNvdXJjZSBjb2RlIGZvciBzb21lIGUtbWFpbCBzeXN0ZW1zLg0KXHBhciANClxw YXIgfXtcYlxmMTExXGZzMjAgRXVkb3JhDQpccGFyIH17XGYxMTFcZnMyMCANClxwYXIgVGhp cyBkb2VzIG5vdCBhcHBlYXIgdG8gYmUgYWJsZSB0byBkaXNwbGF5IHRoZSBzb3VyY2UgY29k ZSBvZiBhbiBlLW1haWwuDQpccGFyIA0KXHBhciB9e1xiXGYxMTFcZnMyMCBOZXRzY2FwZSBh bmQgTW96aWxsYQ0KXHBhciB9e1xmMTExXGZzMjAgDQpccGFyIENsaWNrIFxsZGJscXVvdGUg Vmlld1xyZGJscXVvdGUgIG1lbnUsIGFuZCBcbGRibHF1b3RlIFBhZ2UgU291cmNlXHJkYmxx dW90ZSAuDQpccGFyIA0KXHBhciB9e1xiXGYxMTFcZnMyMCBPcGVyYQ0KXHBhciB9e1xmMTEx XGZzMjAgDQpccGFyIENsaWNrIFxsZGJscXVvdGUgVmlld1xsZGJscXVvdGUgIG1lbnUsIGFu ZCBcbGRibHF1b3RlIFNvdXJjZVxyZGJscXVvdGUgLg0KXHBhciANClxwYXIgfXtcYlxmMTEx XGZzMjAgT3V0bG9vayBFeHByZXNzDQpccGFyIH17XGYxMTFcZnMyMCANClxwYXIgQ2xpY2sg XGxkYmxxdW90ZSBGaWxlXHJkYmxxdW90ZSAgbWVudSwgXGxkYmxxdW90ZSBQcm9wZXJ0aWVz XHJkYmxxdW90ZSAgYW5kIHRoZW4gc2VsZWN0IFxsZGJscXVvdGUgRGV0YWlsc1xyZGJscXVv dGUgIHRhYi4gDQpccGFyIA0KXHBhciB9e1xiXGYxMTFcZnMyMCBQZWdhc3VzDQpccGFyIH17 XGYxMTFcZnMyMCANClxwYXIgVGhpcyBkb2VzIG5vdCBhcHBlYXIgdG8gYmUgYWJsZSB0byBk aXNwbGF5IHRoZSBzb3VyY2UgY29kZSBvZiBhbiBlLW1haWwuDQpccGFyIA0KXHBhciANClxw YXIgDQpccGFyIA0KXHBhciANClxwYXIgDQpccGFyIA0KXHBhciANClxwYXIgfXtcYlxmMTEx IDguIEdsb3NzYXJ5DQpccGFyIH17XGYxMTFcZnMyMCANClxwYXIgQVNDSUkgICAgICAgQW1l cmljYW4gU3RhbmRhcmQgQ29kZSBmb3IgSW5mb3JtYXRpb24gSW50ZXJjaGFuZ2UgDQpccGFy IA0KXHBhciBCaXQgICAgICAgICBUaGUgc21hbGxlc3QgcGFydCBvZiBhIGJ5dGUgd2hpY2gg Y2FuIGJlIGVpdGhlciAwIG9yIDEuDQpccGFyIA0KXHBhciBCeXRlICAgICAgICBBIGNvbXB1 dGVyIGNoYXJhY3RlciBjb25zaXN0aW5nIG9mIDggYml0cyB3aGljaCBjYW4gYmUgdXNlZCB0 byAgDQpccGFyICAgICAgICAgICAgIHJlcHJlc2VudCBhbnkgdmFsdWUgYmV0d2VlbiAwIGFu ZCAyNTUgKDI1NiB2YWx1ZXMgaW4gYWxsKS4NClxwYXIgDQpccGFyIEhleGFkZWNpbWFsIEEg cmVwcmVzZW50YXRpb24gb2YgYSBudW1iZXIgdXNpbmcgYmFzZSAxNiBpbnN0ZWFkIG9mIHRo ZSB1c3VhbA0KXHBhciAgICAgICAgICAgICBiYXNlIDEwLiBJbiB0aGlzIHJlcHJlc2VudGF0 aW9uIHRoZSBkaWdpdHMgYXJlIDAtOSBhbmQgQS1GICgxNg0KXHBhciAgICAgICAgICAgICBp biBhbGwpLCBzbywgZm9yIGV4YW1wbGUsIEhleC4xMCA9IDE2LCAyMCA9IDMyLCAwQSA9IDEw LCBBMCA9IA0KXHBhciAgICAgICAgICAgICAxNjAsIEEzID0gMTYzIGV0Yy4gd2l0aCBhIG1h eGltdW0gaW4gdGhpcyBjYXNlIHdpdGggb25seSB0d28gDQpccGFyICAgICAgICAgICAgIGNo YXJhY3RlcnMsIG9mIEhleC5GRiA9IDI1NS4NClxwYXIgDQpccGFyIElTTyAgICAgICAgIElu dGVybmF0aW9uYWwgU3RhbmRhcmRzIE9yZ2FuaXNhdGlvbiAoaHR0cDovL3d3dy5pc28uY2gp DQpccGFyIA0KXHBhciBNSU1FICAgICAgICBNdWx0aS1wdXJwb3NlIEludGVybmV0IE1haWwg RXh0ZW5zaW9ucw0KXHBhciAgICAgICAgICAgICAoaHR0cDovL3d3dy51ZmFxLm9yZy9uYXZj b20vbWltZV90dXRvcmlhbC5odG1sKQ0KXHBhciANClxwYXIgSWYgeW91IGhhdmUgYSBwcm9i bGVtIHdpdGggYW55IG9mIHRoZSBhYm92ZSBJIHdvdWxkIGJlIHBsZWFzZWQgdG8gaGVhciBm cm9tIHlvdS4NClxwYXIgDQpccGFyIFBsZWFzZSBub3RlIHRoYXQgdGhpcyBmaWxlIGlzIGJl c3Qgdmlld2VkIGluIENvdXJpZXIgZm9udC4NClxwYXIgDQpccGFyIERhdmlkIFdpZ2cNClxw YXIgfVxwYXJkIFxub3dpZGN0bHBhciB7XGYxMTFcZnMyMCB3aWdnamRAYmNzLm9yZy51aw0K XHBhciBNVCBvbiB0aGUgTkVUIFByb2plY3QNClxwYXIgTmF0dXJhbCBMYW5ndWFnZSBUcmFu c2xhdGlvbiBTcGVjaWFsaXN0IEdyb3VwDQpccGFyIFRoZSBCcml0aXNoIENvbXB1dGVyIFNv Y2lldHkNClxwYXIgaHR0cDovL3d3dy5iY3Mub3JnLnVrL3NpZ2dyb3VwL25hbGF0cmFuDQpc cGFyIH19 --------------AAAA09CD11DCC252FCDAF152-- From tpederse@d.umn.edu Mon Feb 3 19:29:33 2003 From: tpederse@d.umn.edu (ted pedersen) Date: Mon, 3 Feb 2003 13:29:33 -0600 (CST) Subject: [MT-List] CFP: HLT-NAACL-03 Workshop on Parallel Text Message-ID: <200302031929.h13JTXW10004@csdev04.d.umn.edu> ** Apologies for Multiple Postings! ** -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- C A L L F O R P A P E R S Building and Using Parallel Texts: Data Driven Machine Translation and Beyond An HLT-NAACL 2003 Workshop Edmonton, Alberta May 31 or June 1, 2003 http://www.cs.unt.edu/~rada/wpt -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- The goal of this workshop is to provide a forum for researchers working on problems related to the creation and use of parallel text. Recent events have demonstrated once again the importance of inter-language communication, and reinforce the need for advances in machine translation (MT) and multi-lingual processing tools. The workshop will be centered around the problem of building and using parallel corpora, which are vital resources for efficiently deriving multi-lingual text processing tools. In addition to regular papers, the workshop also includes a shared task that will result in a comparative evaluation of word alignment techniques. We invite submissions of papers addressing any of the following issues: - Construction of parallel corpora, including the automatic identification and harvesting of parallel corpora from the Web. - Methods to evaluate the quality of parallel corpora and word alignments - Tools for processing parallel corpora, including automatic sentence alignment, word alignment, phrase alignment, detection of omissions and gaps in translations, and others - Using parallel corpora for data driven Machine Translation - Using parallel corpora for the derivation of language processing tools in new languages - Using parallel corpora for automatic corpora annotation - Language learning applied to parallel corpora - Translation memory systems as a source of aligned corpora While we invite submissions addressing any of the above topics, or related issues, we particularly welcome work involving parallel corpora addressing languages with scarce resources. We expect to make arrangements with a journal in Natural Language Processing or Computational Linguistics for a special issue that will include selected papers from this workshop. Invited Speaker: -=-=-=-=-=-=-=-= Elliot Macklovitch, University of Montreal Shared Task: -=-=-=-=-=-= All researchers who have a word alignment system available are invited to participate in the shared task, individually or as part of a team. Participants in the shared task will be provided with common sets of training data, consisting of Romanian-English and French-English parallel texts. Participants will be given approximately one month to train their systems with this data, and then previously held out test data will be released. Participants will run their alignment system on this test data and submit their results, which will be evaluated using a common set of metrics. See the workshop website for details regarding the shared task. Submission format: -=-=-=-=-=-=-=-=-= Submissions should consist of regular full papers of max. 7 pages, formatted following the NAACL 2003 guidelines. In addition, teams participating in the word alignment shared task are invited to submit short papers (max. 4 pages) describing their systems and/or evaluation methodology. Send your submission (a ps or pdf file), prepared for anonymous review, to both: Rada Mihalcea, University of North Texas, rada@cs.unt.edu and Ted Pedersen, University of Minnesota, Duluth, tpederse@d.umn.edu Important dates: -=-=-=-=-=-=-=-= Deadline for regular paper submissions: March 10 Deadline for results submissions: March 25 (shared task) Deadline for short paper submissions: April 1 (shared task) Notification of acceptance for regular papers: April 1 Deadline for camera-ready papers: April 10 Organisation Committee: -=-=-=-=-=-=-=-=-=-=-=- Rada Mihalcea, University of North Texas Ted Pedersen, University of Minnesota, Duluth Program Committee: -=-=-=-=-=-=-=-=-= Lars Ahrenberg, Linkoping University Nicoletta Calzolari, University of Pisa Tim Chklovski, Massachusetts Institute of Technology Mona Diab, University of Maryland Ulrich Germann, Information Sciences Institute Daniel Gildea, University of Pennsylvania Maria das Gracas Volpe Nunes, University of Sao Paulo Nancy Ide, Vassar College Lucia Helena Machado Rino, Federal University of Sao Carlos Eduard Hovy, University of Southern California / Information Sciences Institute Philippe Langlais, University of Montreal Elliot Macklovitch, University of Montreal Daniel Marcu, University of Southern California / Information Sciences Institute Dan Melamed, New York University Magnus Merkel, Linkoping University Ruslan Mitkov, University of Wolverhampton Hermann Ney, RWTH Aachen Franz Och, Information Sciences Institute Kemal Oflazer, Sabanci University Kishore Papineni, IBM Jessie Pinkham, Microsoft Research Andrei Popescu-Belis, ISSCO/TIM/ETI University of Geneva Florence Reeder, MITRE Philip Resnik, University of Maryland Antonio Ribeiro, Joint Research Centre, Ispra, Italy Michel Simard, University of Montreal Harold Somers, University of Manchester Institute of Science and Technology Arturo Trujillo, Canon Research Centre Europe Jean Veronis, University of Provence Clare Voss, Army Research Lab Yorick Wilks, University of Sheffield From tpederse@d.umn.edu Tue Feb 11 18:19:06 2003 From: tpederse@d.umn.edu (ted pedersen) Date: Tue, 11 Feb 2003 12:19:06 -0600 (CST) Subject: [MT-List] Improvements to Latin American Registry Message-ID: <200302111819.h1BIJ6L10197@csdev04.d.umn.edu> There have been some recent improvements to the Latin American Registry for Researchers in Natural Language Processing and Computational Linguistics. http://www.d.umn.edu/~tpederse/registry/registry.cgi Perhaps the main new feature is the addition of a collaborator bulletin board. This allows those located in the rest of the world to make known their interest in working with researchers in Latin America, and vice versa. And of course it can be used to foster colloboration within Latin America too. As of Feb 11, 2003 there are over 80 researchers registered from Argentina, Brazil, Chile, Colombia, Cuba, Ecuador, Mexico, Peru, Uruguay, and Venezuela! It's quite interesting to browse around. Please stop by! Cordially, Ted -- # Ted Pedersen http://www.umn.edu/~tpederse # # Department of Computer Science tpederse@umn.edu # # University of Minnesota, Duluth # # Duluth, MN 55812 (218) 726-8770 # From jyang@systransoft.com Tue Feb 11 21:24:43 2003 From: jyang@systransoft.com (Jin Yang) Date: Tue, 11 Feb 2003 13:24:43 -0800 Subject: [MT-List] MT Summit IX: Call for Workshop Proposals Message-ID: To whom it may concern: Could you post the following at your earliest convenience? Thank you. webmaster@amtaweb.org *** Call for Workshop Proposals *** *** MT Summit IX *** New Orleans, LA, USA, September 23-28, 2003 http://www.mt-summit.org Proposals for workshops are now being solicited on topics of direct interest and impact for MT researchers, developers, vendors or users of MT technologies. Workshops should bring together groups of people involved in a specific problem area of machine translation, to advance the state of the art in that area. Workshops will be held on Saturday 28th September, and will normally last a whole day (approx. 7 hours). Workshops can take a variety of forms, including refereed and invited papers, or other activities. Proposers will be asked to issue their own “Call for Papers”, and to maintain a web site giving information about the submission procedure and, later on, the full programme. Working language of the workshops will normally be English; however, it is acceptable to have a second (or alternative) working language if this is appropriate to the theme of the workshop. This should be stated clearly on the proposal. Submissions should be made to the Workshop Chair (e-mail: Harold.Somers@umist.ac.uk) by March 21st. They should include the theme and goal of the workshop, the planned activities, a calendar of deadlines for submission, notification and, if appropriate, camera-ready copy (we advise that you follow the submission dates for the main conference), and an estimate of the number of participants. Important Dates 21 March 2003 Deadline for Workshop proposals 31 March 2003 Notification of acceptance of proposal 11 May 2003 Main conference paper submission deadline 30 June 2003 Main conference paper notification 31 July 2003 Main conference camera-ready copy deadline For general conference information and further details as it becomes available, visit the MT Summit web site http://www.mt-summit.org From jyang@systransoft.com Fri Feb 14 19:28:44 2003 From: jyang@systransoft.com (Jin Yang) Date: Fri, 14 Feb 2003 11:28:44 -0800 Subject: [MT-List] MT Summit IX: Call for Workshop Proposals Message-ID: To whom it may concern: There is a mistake in the previous "Call for Workshop Proposals". Could you repost the following corrected one? Best regards, webmaster@amtaweb.org *** Call for Workshop Proposals *** *** MT Summit IX *** New Orleans, LA, USA, September 23-28, 2003 http://www.mt-summit.org Proposals for workshops are now being solicited on topics of direct interest and impact for MT researchers, developers, vendors or users of MT technologies. Workshops should bring together groups of people involved in a specific problem area of machine translation, to advance the state of the art in that area. Workshops will be held on ***Saturday 27th September***, and will normally last a whole day (approx. 7 hours). Workshops can take a variety of forms, including refereed and invited papers, or other activities. Proposers will be asked to issue their own “Call for Papers”, and to maintain a web site giving information about the submission procedure and, later on, the full programme. Working language of the workshops will normally be English; however, it is acceptable to have a second (or alternative) working language if this is appropriate to the theme of the workshop. This should be stated clearly on the proposal. Submissions should be made to the Workshop Chair (e-mail: Harold.Somers@umist.ac.uk) by March 21st. They should include the theme and goal of the workshop, the planned activities, a calendar of deadlines for submission, notification and, if appropriate, camera-ready copy (we advise that you follow the submission dates for the main conference), and an estimate of the number of participants. Important Dates 21 March 2003 Deadline for Workshop proposals 31 March 2003 Notification of acceptance of proposal 11 May 2003 Main conference paper submission deadline 30 June 2003 Main conference paper notification 31 July 2003 Main conference camera-ready copy deadline For general conference information and further details as it becomes available, visit the MT Summit web site http://www.mt-summit.org From tpederse@d.umn.edu Thu Feb 27 00:33:39 2003 From: tpederse@d.umn.edu (ted pedersen) Date: Wed, 26 Feb 2003 18:33:39 -0600 (CST) Subject: [MT-List] CFP: HLT-NAACL-03 Shared Task on Word Alignment Message-ID: <200302270033.h1R0Xdv14820@csdev11.d.umn.edu> ----------------------------------- CALL FOR PARTICIPATION Shared Task on Word Alignment HLT/NAACL 2003 Workshop "Building and Using Parallel Texts" ------------------------------------ All researchers who have a word alignment system available are invited to participate in this shared task, individually or as part of a team. Participants in the shared task will be invited to submit short papers describing their system and/or evaluation methodology, which will be published in the workshop proceedings. Training and trial data for the two word alignment tasks (Romanian-English and English-French) are already available from the workshop webpage: http://www.cs.unt.edu/~rada/wpt Shared task timetable: Training and trial data available February 14 Test data available March 18 Submission of results March 25 Results back to participants March 28 Submission of short papers April 1 Last day to register for participation in the shared task: March 21. [no more access to training/trial/test data will be given after this date] From roisin_saunier@hotmail.com Mon Mar 10 15:29:42 2003 From: roisin_saunier@hotmail.com (Roisin Saunier) Date: Mon, 10 Mar 2003 16:29:42 +0100 Subject: [MT-List] Cultural Factors and MT Message-ID: Dear All, Does anyone know of research that has been carried out in relation to how cultural factors influence: (1) how users interact with an MT system (2) users' acceptance or rejection of MT technology (3) the acceptability of MT output >From my own observations of and discussions with end users, their cultural backgrounds do indeed seem to affect the above. So I'm interested in the results of studies that would either substantiate or disprove my 'theory'. Many thanks for any references you can supply, Róisín _________________________________________________________________ Add photos to your messages with MSN 8. Get 2 months FREE*. http://join.msn.com/?page=features/featuredemail From jeff.allen@free.fr Fri Mar 14 21:27:31 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Fri, 14 Mar 2003 22:27:31 +0100 Subject: [MT-List] EAMT/CLAW2003 Registration is now open Message-ID: <044201c2ea73$2f15ed60$f199933e@home> *************** Registration now Open *************** Joint Conference combining the 7th International Workshop of the European Association for Machine Translation (EAMT) and the 4th Controlled Language Applications Workshop (CLAW) Main Conference theme: Controlled Language Translation Location: Dublin City University, Ireland Dates: 15-17 May, 2003 General Conference URL: http://www.eamt.org/eamt-claw03/ Registration URL: http://www.eamt.org/eamt-claw03/register.html Closing date for early registration: 31 March 2003 NOTE: in order to benefit from the special accommodation rates, you must book your room BEFORE the 1st of April, 2003. Best regards, Jeff Allen From WJHutchins@compuserve.com Mon Mar 17 10:30:46 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Mon, 17 Mar 2003 05:30:46 -0500 Subject: [MT-List] Re: EAMT workshop at EACL 03 Message-ID: <200303170531_MC3-1-2FB9-E9BD@compuserve.com> Reminder! EAMT is holding a one-day workshop at the EACL conference in Budapest, April 12-17. For details see: http://www.elsnet.org/workshops-eacl2003/index.html Deadline for early registration has now passed. John Hutchins= From tpederse@d.umn.edu Wed Mar 5 14:55:48 2003 From: tpederse@d.umn.edu (ted pedersen) Date: Wed, 5 Mar 2003 08:55:48 -0600 (CST) Subject: [MT-List] FINAL CFP: HLT-NAACL-03 Workshop on Parallel Text Message-ID: <200303051455.h25EtmR11684@csdev09.d.umn.edu> == Apologies for multiple postings == Notice the March 10 Deadline for regular papers submissions! -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- C A L L F O R P A P E R S Building and Using Parallel Texts: Data Driven Machine Translation and Beyond An HLT-NAACL 2003 Workshop Edmonton, Alberta May 31, 2003 http://www.cs.unt.edu/~rada/wpt -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- The goal of this workshop is to provide a forum for researchers working on problems related to the creation and use of parallel text. Recent events have demonstrated once again the importance of inter-language communication, and reinforce the need for advances in machine translation (MT) and multi-lingual processing tools. The workshop will be centered around the problem of building and using parallel corpora, which are vital resources for efficiently deriving multi-lingual text processing tools. In addition to regular papers, the workshop also includes a shared task that will result in a comparative evaluation of word alignment techniques. We invite submissions of papers addressing any of the following issues: - Construction of parallel corpora, including the automatic identification and harvesting of parallel corpora from the Web. - Methods to evaluate the quality of parallel corpora and word alignments - Tools for processing parallel corpora, including automatic sentence alignment, word alignment, phrase alignment, detection of omissions and gaps in translations, and others - Using parallel corpora for data driven Machine Translation - Using parallel corpora for the derivation of language processing tools in new languages - Using parallel corpora for automatic corpora annotation - Language learning applied to parallel corpora - Translation memory systems as a source of aligned corpora While we invite submissions addressing any of the above topics, or related issues, we particularly welcome work involving parallel corpora addressing languages with scarce resources. We expect to make arrangements with a journal in Natural Language Processing or Computational Linguistics for a special issue that will include selected papers from this workshop. Invited Speaker: -=-=-=-=-=-=-=-= Elliot Macklovitch, University of Montreal Shared Task: -=-=-=-=-=-= All researchers who have a word alignment system available are invited to participate in the shared task, individually or as part of a team. Participants in the shared task will be provided with common sets of training data, consisting of Romanian-English and French-English parallel texts. Participants will be given approximately one month to train their systems with this data, and then previously held out test data will be released. Participants will run their alignment system on this test data and submit their results, which will be evaluated using a common set of metrics. See the workshop website for details regarding the shared task. Submission format: -=-=-=-=-=-=-=-=-= Submissions should consist of regular full papers of max. 7 pages, formatted following the NAACL 2003 guidelines. In addition, teams participating in the word alignment shared task are invited to submit short papers (max. 4 pages) describing their systems and/or evaluation methodology. Send your submission (a ps or pdf file), prepared for anonymous review, to both: Rada Mihalcea, University of North Texas, rada@cs.unt.edu and Ted Pedersen, University of Minnesota, Duluth, tpederse@d.umn.edu Important dates: -=-=-=-=-=-=-=-= Deadline for regular paper submissions: March 10 Deadline for results submissions: March 25 (shared task) Deadline for short paper submissions: April 1 (shared task) Notification of acceptance for regular papers: April 1 Deadline for camera-ready papers: April 10 Organisation Committee: -=-=-=-=-=-=-=-=-=-=-=- Rada Mihalcea, University of North Texas Ted Pedersen, University of Minnesota, Duluth Program Committee: -=-=-=-=-=-=-=-=-= Lars Ahrenberg, Linkoping University Nicoletta Calzolari, University of Pisa Tim Chklovski, Massachusetts Institute of Technology Mona Diab, University of Maryland Ulrich Germann, Information Sciences Institute Daniel Gildea, University of Pennsylvania Maria das Gracas Volpe Nunes, University of Sao Paulo Nancy Ide, Vassar College Lucia Helena Machado Rino, Federal University of Sao Carlos Eduard Hovy, University of Southern California/Information Sciences Institute Philippe Langlais, University of Montreal Elliot Macklovitch, University of Montreal Daniel Marcu, University of Southern California/Information Sciences Institute Dan Melamed, New York University Magnus Merkel, Linkoping University Ruslan Mitkov, University of Wolverhampton Grace Ngai, Hong Kong Polytechnic University Hermann Ney, RWTH Aachen Franz Och, Information Sciences Institute Kemal Oflazer, Sabanci University Kishore Papineni, IBM Jessie Pinkham, Microsoft Research Andrei Popescu-Belis, ISSCO/TIM/ETI University of Geneva Florence Reeder, MITRE Philip Resnik, University of Maryland Antonio Ribeiro, Joint Research Centre, Ispra, Italy Michel Simard, University of Montreal Harold Somers, University of Manchester Institute of Science and Technology Arturo Trujillo, Canon Research Centre Europe Jean Veronis, University of Provence Clare Voss, Army Research Lab Dan Tufis, RACAI Romania Yorick Wilks, University of Sheffield From jeff.allen@free.fr Fri Mar 14 21:40:12 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Fri, 14 Mar 2003 22:40:12 +0100 Subject: [MT-List] updated links: articles in MLCT on MT, TM, postediting Message-ID: <044301c2ea73$2fa3d580$f199933e@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_0429_01C2EA7A.ACEE4000 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear MT-listers,=20 Please find below updated link information on several articles published = in Multilingual Computing & Technology magazine that pertain to Machine = Translation, Translation Memory, and Post-editing. Several at the end = of the list are reviews of old versions of products, but at least these = are existing published reviews that can be accessed online. =20 If anyone is interested in conducting reviews of MT and TM products for = the magazine, please let me know so that we can arrange it with the = managing editor. (warning: many of the http links below are truncated due to their = length. For such links, you will need to copy and paste all parts of the = entire http path into your web browser. All articles can also be found = by conducting a search on the www.multilingual.com site) =20 Regards, Jeff Allen ------ Multilingual Computing & Technology=20 55 Volume 14 Issue 3 Several articles and reviews on translation software related topics = scheduled to appear in this upcoming issue. Multilingual Computing & Technology=20 54 Volume 14 Issue 2 Conference Combines MT and Controlled Language Topics: European = Association for Machine Translation and Controlled Language Applications = Workshop to meet http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2ffeat uredarticlesdetail.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortord= er=3Ddesc end&-op=3Deq&Ad%20Type=3Dreprint&-max=3D5&-recid=3D33457&-token=3D%5bFMP-= currenttoken% 5d&-find=3D Multilingual Computing & Technology=20 53 Volume 14 Issue 1 Review of Online Bible Millennium Edition 1.11 (software for accessing = multi-reference translation parallel language databases) http://www.multilingual.com/allen53.htm Note: French translation of article at: http://www.editionscle.com/bol/presse/article2/allen-mltc53-fr.htm Multilingual Computing & Technology=20 53 Volume 14 Issue 1 Computing in Creoles (survey of Creole language technologies) http://www.multilingual.com/masonAllen53.htm =20 Multilingual Computing & Technology=20 #52 Volume 13 Issue 8 Review of TRADOS 5.5 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33440& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #51 Volume 13 Issue 7 Review of MultiTrans 3.0 (Corpus-based translation support tool) =20 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33422& -token=3Dnow&-find=3D =20 Multilingual Computing & Technology=20 #51 Volume 13 Issue 7 The Bible as a Resource for Translation Software. Multilingual Computing = and Technology magazine. Number 51, Vol. 13, Issue 7. October/November 2002. Pp. 40-45. http://www.multilingual.com/allen51.htm =20 Note: French translation of article at: http://www.editionscle.com/bol/presse/article1/allen-mltc51-fr.htm =20 Multilingual Computing & Technology=20 #50 Volume 13 Issue 6 Review of Reverso Pro 5 and Reverso Expert http://www.multilingual.com/allen50.htm Multilingual Computing & Technology=20 #50 Volume 13 Issue 6 T-Remote Memory (new framework for sharing translation memory) http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33413& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #50 Volume 13 Issue 6 Working with Machine Translation: how the Euroscript Group uses Systran = for fast news translation. Annette Bus and Anja Simon. 2002 = Multilingual Computing and Technology magazine. #50 Volume 13 Issue 6. = Pp 37-41. search for article online at http://www.multilingual.com/=20 Multilingual Computing & Technology=20 #50 Volume 13 Issue 6 Intregrating Translation tools with document creation: Caterpillar uses = D=E9j=E0 Vu and authoring tools to create localized manuals. Kirsi = Rintanen and Jost Zetzsche. 2002. Multilingual Computing and Technology = magazine. #50 Volume 13 Issue 6. Pp 42-44. search for article online = at http://www.multilingual.com/=20 =20 Multilingual Computing & Technology=20 #47 Volume 13 Issue 3 Review of RC-WinTrans 6.01 (Tool for localizing Windows applications) http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33372& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #47 Volume 13 Issue 3 Review of MetaTexis .0999g =20 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33373& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #47 Volume 13 Issue 3 Systems for evaluating translation quality: SAE J2450 and ITR Blackjack = offer two approaches to ensuring translation consistency. Helen = Eckersley. 2002. In Multilingual Computing and Technology magazine. = Number 47, Vol. 13, Issue 3. April/May 2002. Pp. 39-42. search for = article online at http://www.multilingual.com/=20 =20 Multilingual Computing & Technology=20 #46 Volume 13 Issue 2 =20 Review of Repairing Texts: Empirical Investigations of Machine = Translation Post-Editing Processes. (KRINGS Hans, edited by Geoffrey = KOBY. 2001. Translated from German to English by Geoffrey Koby, Gregory = Shreve, Katjz Mischerikow and Sarah Litzer) Translation Studies series. = Ohio: Kent State University Press. In Multilingual Computing and = Technology magazine. Number 46. March 2002. Pp. 27-29. Review available at: http://www.multilingual.com/allen46.htm Book ordering information at: = http://bookmasters.com/ksu-press/ksu071.htm =20 Multilingual Computing & Technology=20 #46 Volume 13 Issue 2 Translation needs in auto manufacturing. Rick Woyde. 2002. In = Multilingual Computing and Technology magazine. Number 46, Vol. 13, = Issue 2. March 2002. Pp. 39-42. search for article online at = http://www.multilingual.com/ Multilingual Computing & Technology=20 #45 Volume 13 Issue 1 Review of Wordfast (Macintosh version) http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33341& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #45 Volume 13 Issue 1 Review of Wordfast (Windows version)=20 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33342& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #44 Volume 12 Issue 8 Review of Translate =20 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33330& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #29 Volume 11 Issue 1 Review of Desktop Translator 1 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33168& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #28 Volume 10 Issue 6 Review of L&H Power Translator Pro 6.43 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33156& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #27 Volume 10 Issue 5 Review of E-J QuickTransView 1.0, E-J Pro 5.0, E-J Internet Plus http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D33145& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #21 Volume 9 Issue 5 Review of TransLinGo! 1.0 (Japanese to English Web page translation) http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D32842& -token=3Dnow&-find=3D Multilingual Computing & Technology=20 #21 Volume 9 Issue 5 Review of T1 Professional 3.0 http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Dourpublication= %2frevi ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3D= descend &-op=3Deq&Ad%20Type=3Dreview&-op=3Deq&Review%20Type=3Dproduct&-max=3D500&= -recid=3D32873& -token=3Dnow&-find=3D --- end --- ------=_NextPart_000_0429_01C2EA7A.ACEE4000 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Dear MT-listers,
 
Please find below updated link information on = several articles=20 published in Multilingual Computing & Technology magazine that = pertain to=20 Machine Translation, Translation Memory, and Post-editing.  Several = at the=20 end of the list are reviews of old versions of products, but at least = these are=20 existing published reviews that can be accessed online.  =
If anyone is interested in conducting reviews of MT = and TM=20 products for the magazine, please let me know so that we = can arrange=20 it with the managing editor.
 
(warning: many of the http links below are truncated = due to=20 their length. For such links, you will need to copy and paste all = parts of=20 the entire http path into your web browser.  All articles can = also be=20 found by conducting a search on the www.multilingual.com = site)

 
Regards,
 
Jeff Allen
 
------
 
Multilingual = Computing &=20 Technology
55 Volume 14 Issue 3
 
Several articles and reviews on translation software = related=20 topics scheduled to appear in this upcoming issue.
 
 
Multilingual Computing & = Technology
54=20 Volume 14 Issue 2
 
Conference Combines MT and Controlled Language = Topics:=20 European Association for Machine Translation and Controlled Language=20 Applications Workshop to meet
 
http://www.multilingual.com/FMPro?-db=3Darchives&-for= mat=3Dourpublication%2ffeat
uredarticlesdetail.htm&-lay= =3Dcgi&-sortfield=3Dmagazine%20number&-sortorder=3Ddesc
end&am= p;-op=3Deq&Ad%20Type=3Dreprint&-max=3D5&-recid=3D33457&-t= oken=3D%5bFMP-currenttoken%
5d&-find=3D
 
 
Multilingual Computing & = Technology
53=20 Volume 14 Issue 1
 
Review of Online Bible Millennium Edition 1.11 = (software for=20 accessing multi-reference translation parallel language = databases)
 
 
Note: French = translation of article=20 at:
http://www.editionscle.com/bol/presse/article2/allen-mltc= 53-fr.htm
 
 
Multilingual = Computing &=20 Technology
53 Volume 14 Issue 1
 

Computing in Creoles (survey of Creole language=20 technologies)

http://www.multilingual.com/masonAllen53.htm
 
 
 
Multilingual Computing & = Technology
#52=20 Volume 13 Issue 8
 
Review of TRADOS 5.5

http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33440&= amp;
-token=3Dnow&-find=3D
 
 
Multilingual Computing & = Technology
#51=20 Volume 13 Issue 7
 
Review of MultiTrans 3.0 (Corpus-based translation = support=20 tool)

 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33422&= amp;
-token=3Dnow&-find=3D
 
 =20
Multilingual Computing & = Technology
#51=20 Volume 13 Issue 7
 
The Bible as a Resource for Translation Software. = Multilingual=20 Computing and Technology magazine. Number 51, Vol. 13, Issue=20 7.
October/November 2002. Pp. 40-45.
http://www.multilingual.com/allen51.htm
 
Note: French translation of article = at:
http://www.editionscle.com/bol/presse/article1/allen-mltc= 51-fr.htm
 
 
Multilingual Computing & = Technology=20
#50 Volume 13 Issue 6
 
Review of Reverso Pro 5 and Reverso = Expert
http://www.multilingual.com/allen50.htm
 
 
 
Multilingual Computing & = Technology
#50=20 Volume 13 Issue 6
 
T-Remote Memory (new framework for sharing = translation=20 memory)
 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33413&= amp;
-token=3Dnow&-find=3D
 
 
Multilingual = Computing &=20 Technology
#50 Volume 13 Issue 6
 
Working with Machine Translation: how the Euroscript = Group=20 uses Systran for fast news translation. Annette Bus and Anja Simon. = 2002 =20 Multilingual Computing and Technology magazine. #50 Volume 13 Issue = 6.  Pp=20 37-41.  search for article online at http://www.multilingual.com/
 
 
Multilingual Computing & = Technology
#50=20 Volume 13 Issue 6
 
Intregrating Translation tools with document = creation:=20 Caterpillar uses D=E9j=E0 Vu and authoring tools to create localized = manuals. Kirsi=20 Rintanen and Jost Zetzsche. 2002.  Multilingual Computing and = Technology=20 magazine. #50 Volume 13 Issue 6.  Pp 42-44.  search for = article online=20 at http://www.multilingual.com/
 
 
 
Multilingual Computing & = Technology
#47=20 Volume 13 Issue 3
 
Review of RC-WinTrans 6.01 (Tool for localizing = Windows=20 applications)
 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33372&= amp;
-token=3Dnow&-find=3D
 
 
Multilingual Computing & = Technology
#47=20 Volume 13 Issue 3
 
Review of MetaTexis .0999g

 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33373&= amp;
-token=3Dnow&-find=3D
 
 
Multilingual Computing & = Technology
#47=20 Volume 13 Issue 3
 
Systems for evaluating translation quality: SAE = J2450 and ITR=20 Blackjack offer two approaches to ensuring translation consistency. = Helen=20 Eckersley. 2002. In Multilingual Computing and Technology magazine. = Number 47,=20 Vol. 13, Issue 3. April/May 2002. Pp. 39-42. search for article online = at http://www.multilingual.com/ =
 
 
Multilingual = Computing &=20 Technology
#46 = Volume 13=20 Issue 2

 
Review of Repairing Texts: Empirical Investigations = of Machine=20 Translation Post-Editing Processes. (KRINGS Hans, edited by Geoffrey = KOBY. 2001.=20 Translated from German to English by Geoffrey Koby, Gregory Shreve, = Katjz=20 Mischerikow and Sarah Litzer) Translation Studies series. Ohio: Kent = State=20 University Press. In Multilingual Computing and Technology magazine. = Number 46.=20 March 2002. Pp. 27-29.
Review available at: http://www.multilingual.com/allen46.htm
 
Book ordering information at: http://bookmasters.com/ksu-press/ksu071.htm
 
 
Multilingual=20 Computing & Technology
#46 Volume = 13 Issue=20 2
 
Translation needs in auto manufacturing. Rick Woyde. = 2002. In=20 Multilingual Computing and Technology magazine. Number 46, Vol. 13, = Issue 2.=20 March 2002. Pp. 39-42. search for article online at http://www.multilingual.com/
 
 
Multilingual Computing & Technology=20
#45 Volume 13 Issue=20 1
 
Review of Wordfast (Macintosh version)

http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33341&= amp;
-token=3Dnow&-find=3D
 

Multilingual Computing & Technology=20
#45 Volume 13 Issue=20 1
 
Review of Wordfast (Windows version)
 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33342&= amp;
-token=3Dnow&-find=3D
 
 
Multilingual Computing & = Technology=20
#44 Volume = 12 Issue=20 8
 
Review of Translate
 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33330&= amp;
-token=3Dnow&-find=3D
 
Multilingual = Computing &=20 Technology
#29=20 Volume 11 Issue 1
 
Review of Desktop Translator 1
 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33168&= amp;
-token=3Dnow&-find=3D
 
Multilingual = Computing &=20 Technology
#28=20 Volume 10 Issue 6
 
Review of L&H Power Translator Pro = 6.43
 
http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33156&= amp;
-token=3Dnow&-find=3D
 
Multilingual = Computing &=20 Technology
#27=20 Volume 10 Issue 5
 
Review of E-J QuickTransView 1.0, E-J Pro 5.0, E-J = Internet=20 Plus

http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D33145&= amp;
-token=3Dnow&-find=3D
 
Multilingual = Computing &=20 Technology
#21=20 Volume 9 Issue 5
 
Review of TransLinGo! 1.0 (Japanese to English Web = page=20 translation)

http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D32842&= amp;
-token=3Dnow&-find=3D
 
 
Multilingual = Computing &=20 Technology
#21=20 Volume 9 Issue 5
 
Review of T1 Professional 3.0

http://www.multilingual.com/FMPro?-db=3Darchives&-format=3Do= urpublication%2frevi
ewdetailproduct.htm&-lay=3Dcgi&-sortfield=3Dmagazine%20n= umber&-sortorder=3Ddescend
&-op=3Deq&Ad%20Type=3Dreview&am= p;-op=3Deq&Review%20Type=3Dproduct&-max=3D500&-recid=3D32873&= amp;
-token=3Dnow&-find=3D
 
 
 
--- end ---
------=_NextPart_000_0429_01C2EA7A.ACEE4000-- From jeff.allen@free.fr Sun Mar 23 08:10:42 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Sun, 23 Mar 2003 09:10:42 +0100 Subject: [MT-List] additional article in MLCT on MT References: <044301c2ea73$2fa3d580$f199933e@home> Message-ID: <001901c2f171$d7c4e9c0$cf9c933e@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_0318_01C2F11C.1490AFC0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable My apologies for forgetting to mention the following article in my last = posting. Multilingual Computing & Technology=20 #47 Volume 13 Issue 3 An Overview of PROMT Translation Software. Svetlana Sokolova. 2002. In = Multilingual Computing and Technology magazine. Number 47, Vol. 13, = Issue 3. April/May 2002. Pp. 43-46. search for article online at = http://www.multilingual.com/=20 =20 Regards, =20 Jeff Allen =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Friday, March 14, 2003 10:40 PM Objet : [MT-List] updated links: articles in MLCT on MT, TM, = postediting Dear MT-listers,=20 Please find below updated link information on several articles = published in Multilingual Computing & Technology magazine that pertain = to Machine Translation, Translation Memory, and Post-editing. Several = at the end of the list are reviews of old versions of products, but at = least these are existing published reviews that can be accessed online. = If anyone is interested in conducting reviews of MT and TM products = for the magazine, please let me know so that we can arrange it with the = managing editor. (warning: many of the http links below are truncated due to their = length. For such links, you will need to copy and paste all parts of the = entire http path into your web browser. All articles can also be found = by conducting a search on the www.multilingual.com site) =20 Regards, Jeff Allen ------=_NextPart_000_0318_01C2F11C.1490AFC0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
My apologies for forgetting to mention the following = article=20 in my last posting.
 
Multilingual Computing & = Technology
#47=20 Volume 13 Issue 3
 
An Overview of PROMT Translation Software. Svetlana = Sokolova.=20 2002. In Multilingual Computing and Technology magazine. Number 47, Vol. = 13,=20 Issue 3. April/May 2002. Pp. 43-46. search for article online at http://www.multilingual.com/
 
Regards,
 
Jeff Allen
 
 
----- Message d'origine -----
De=20 : Jeff=20 Allen
Envoy=E9 : Friday, March = 14, 2003=20 10:40 PM
Objet : [MT-List] updated = links: articles=20 in MLCT on MT, TM, postediting

Dear MT-listers,
 
Please find below updated link information on = several=20 articles published in Multilingual Computing & Technology magazine = that=20 pertain to Machine Translation, Translation Memory, and = Post-editing. =20 Several at the end of the list are reviews of old versions of = products, but at=20 least these are existing published reviews that can be accessed = online. =20
If anyone is interested in conducting reviews of = MT and TM=20 products for the magazine, please let me know so that we = can arrange=20 it with the managing editor.
 
(warning: many of the http links below are = truncated due to=20 their length. For such links, you will need to copy and paste all = parts=20 of the entire http path into your web browser.  All articles = can=20 also be found by conducting a search on the www.multilingual.com = site)

 
Regards,
 
Jeff Allen
 
<snip>
 
------=_NextPart_000_0318_01C2F11C.1490AFC0-- From jyang@systransoft.com Wed Mar 26 23:35:10 2003 From: jyang@systransoft.com (Jin Yang) Date: Wed, 26 Mar 2003 15:35:10 -0800 Subject: [MT-List] updated CFP for MT Summit IX Message-ID: To whom it may concern: Below please find the updated CFP for MT Summit IX. Please distribute the CFP at your convenience. Thank you. webmaster@amtaweb.org MACHINE TRANSLATION SUMMIT IX September 23-28, 2003 New Orleans, USA http://www.mt-summit.org CALL FOR PAPERS The ninth Machine Translation Summit, organized by the International Association for Machine Translation (IAMT) and hosted by the Association for Machine Translation in the Americas (AMTA), will be held in New Orleans, Louisiana, from 23 to 28 September 2003. MT Summit IX hereby invites original submissions on all aspects machine and machine-aided translation. Please consult the conference Web site for the complete Call for Papers, including details on the different categories of submissions. Important dates: - May 11, 2003 : Deadline for the submission of papers - June 30, 2003 : Notification of acceptance sent to authors - July 31, 2003 : Camera-ready copy due - September 23, 2003 : Conference opens Invited Speakers: MT Summit IX will feature invited talks by the following speakers: Pierre Isabelle Area Manager, Content Analysis, Xerox Research Centre Europe, Grenoble "Multilingual Document Processing at XRCE" Akitoshi Okumura Senior Manager, Human Language Technology Group NEC Corporation, Japan "Development of Speech Translation for Hand-held Devices" Donald Barabé Director, Business Development Translation Bureau, Public Works and Government Services Canada "Soaring Demand, Shrinking Supply in Translation: How We Plan to Make Ends Meet" From jeff.allen@free.fr Wed Apr 2 05:47:32 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Wed, 2 Apr 2003 07:47:32 +0200 Subject: [MT-List] more articles in MLCT on MT and translation technologies Message-ID: <00da01c2f8df$d1e47140$0100007f@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_0012_01C2F8EC.1E24E560 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear all, Still a few other articles (including 1 review) on MT and translation = related technologies that were not included in my last posting on this = topic. Regards, Jeff Allen Multilingual Computing & Technology=20 54 Volume 14 Issue 2 TMX Implementation in Major Translation Tools=20 Jost Zetzsche. In Multilingual Computing and Technology magazine. Number = 54, Vol. 14, Issue 2. March 2003. Pp. 23-27. This article includes interviews with the following people and topics: Segmentation Algorithms and 100% Matches (Philippe Mercier) TMX and Its Role in the Marketplace (Hedley Rees-Evans) SMGL and XML in a Multilingual Information Process (Hedley Rees-Evans) search for article online at http://www.multilingual.com/=20 Multilingual Computing & Technology=20 53 Volume 14 Issue 1 Review of Personal Translator 2002 Office Plus. =20 G. Daniel Bugel-Shunra. In Multilingual Computing and Technology = magazine. Number 53, Vol. 14, Issue 1. January/February 2003. Pp. 19-20. search for article online at http://www.multilingual.com/=20 Multilingual Computing & Technology=20 #51 Volume 13 Issue 7 Challenges in Arabic Natural Langauge Processing Achraf Chalabi. 2002 In Multilingual Computing and Technology magazine. = #51 Volume 13 Issue 7. Pp 35-36. search for article online at = http://www.multilingual.com/=20 =20 Multilingual Computing & Technology=20 #51 supplement (October/November 2002) on Language Technology=20 Getting Started guide series =20 What is Language Technology? by Chris Langewis. =20 Comparing Basic Features of TM tools by Angelika Zerfass Technology and the Freelance Translator Jonathan T. Hine, Jr. Finding the Right Language Technology Tools by David Shadbolt search for articles online at http://www.multilingual.com/=20 =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Friday, March 14, 2003 10:40 PM Objet : [MT-List] updated links: articles in MLCT on MT, TM, postediting ------=_NextPart_000_0012_01C2F8EC.1E24E560 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Dear all,
 
Still a few other articles (including 1 review) on MT and = translation=20 related technologies that were not included in my last posting on this=20 topic.
 
Regards,
 
Jeff Allen
 
 
 
Multilingual Computing & = Technology
54=20 Volume 14 Issue 2
 
TMX Implementation in Major Translation=20 Tools 
Jost Zetzsche. In Multilingual = Computing=20 and Technology magazine. Number 54, Vol. 14, Issue 2. March 2003. = Pp.=20 23-27.
This article includes interviews with the following people and=20 topics:
Segmentation Algorithms and 100% Matches (Philippe Mercier)
TMX and Its Role in the Marketplace (Hedley Rees-Evans)
SMGL and XML in a Multilingual Information Process (Hedley=20 Rees-Evans)
 
search for article online at http://www.multilingual.com/ 
 
 
Multilingual Computing & = Technology
53=20 Volume 14 Issue 1
 
Review of Personal Translator 2002 Office=20 Plus.  
G. Daniel Bugel-Shunra. In Multilingual = Computing and=20 Technology magazine. Number 53, Vol. 14, Issue 1. = January/February 2003.=20 Pp. 19-20.
search for article online at http://www.multilingual.com/ 
 
Multilingual Computing & = Technology
#51=20 Volume 13 Issue 7
 
Challenges in Arabic Natural Langauge = Processing
Achraf Chalabi. 2002  In Multilingual Computing = and=20 Technology magazine. #51 Volume 13 Issue 7.  Pp 35-36.  search = for=20 article online at http://www.multilingual.com/ 
<= /DIV>
 
Multilingual Computing & = Technology
#51=20 supplement (October/November 2002) on Language Technology
Getting Started guide=20 series
 
What is Language Technology?
by Chris Langewis.
 
Comparing Basic Features of TM tools
by Angelika Zerfass
 
Technology and the Freelance Translator
Jonathan T. Hine, Jr.
 
Finding the Right Language Technology Tools
by David Shadbolt
 
search for articles online at http://www.multilingual.com/ 
 
 
  ----- = Message=20 d'origine -----
De : Jeff = Allen
Envoy=E9 : Friday, March 14, 2003 10:40 PM
Objet : [MT-List] updated links: articles in MLCT on MT, TM, = postediting

<snip>
------=_NextPart_000_0012_01C2F8EC.1E24E560-- From jeff.allen@free.fr Wed Apr 2 05:56:28 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Wed, 2 Apr 2003 07:56:28 +0200 Subject: [MT-List] contacting Karin Spalink References: Message-ID: <00de01c2f8df$e07d51e0$0100007f@home> Does anyone have recent contact info for Karin Spalink? The e-mail address I used during mid-2002 to correspond with her is now bouncing. Please reply to me directly. Thanks, Jeff --------------------------------- Jeff ALLEN Paris, FRANCE e-mail: jeff.allen@free.fr From WJHutchins@compuserve.com Thu Apr 3 10:26:59 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Thu, 3 Apr 2003 05:26:59 -0500 Subject: [MT-List] new edition of Compendium Message-ID: <200304030527_MC3-1-3290-19D5@compuserve.com> The latest edition of the "Compendium of translation software" (March 2003) is now available on the EAMT website: http://www.eamt.org/compendium.html If you see any errors or omissions please let me know, and I shall include changes in the next update due in August. For those interested in seeing what was available in the past there are back editions (PDF) available on my website: http://ourworld.compuserve.com/homepages/WJHutchins John Hutchins 3 April From jeff.allen@free.fr Tue Apr 8 10:26:23 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Tue, 8 Apr 2003 11:26:23 +0200 Subject: [MT-List] seeking human translation statistics References: <200304030527_MC3-1-3290-19D5@compuserve.com> Message-ID: <00c801c2fdb0$f50f7ea0$1098933e@home> Dear all, I am looking for published statements about the speed at which human translators and translation reviewers conduct their tasks. The types of statements I seek would resemble the following: * "Typically, a translator is able to translate between X and Y words per day, depending on the type and technical difficulty of the document" * "Their translation speed is about X pages or Y words per day " * "Our statistics show that translators for X pair of languages translate Y number of words per day" * etc. I have collected a few examples of such quotes from books, articles, and web sites, yet I am looking for obtainable and citable examples from as many sources as possible. Such cross-validation will help verify statements that are made on translation speed. If you are willing to share such quotes and references, please contact me with the information at my e-mail below. I will share the set of quotable references in an upcoming article, and will acknowledge those who participate in providing such information. Regards, Jeff Allen --------------------------------- Jeff ALLEN Paris, France Mobile Tel: (+33) 6 86 06 87 59 e-mail: jeff.allen@free.fr or postediting@hotmail.com From jeff.allen@free.fr Tue Apr 8 10:27:17 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Tue, 8 Apr 2003 11:27:17 +0200 Subject: [MT-List] Published translation worse than raw MToutput References: Message-ID: <00d101c2fdb1$0f264440$1098933e@home> Dear all, I recently purchased a scooter board which came with use and maintenance instructions in several languages. Below is "exactly" how the French and English texts appear in these instructions. If you send the French text below through your favorite online MT system, it is very likely that the raw MT English output in many sentences will be better than the published English text that came with the item. At least this is what I found when I ran the French text through a system or two. Regards, Jeff Allen ------- begin ----- NOTICE A CONSERVER CHARGE MINIMALE D'UTILISATION: 20 KG ATTENTION: A UTILISER AVEC DES EQUIPEMENTS DE PROTECTION (CASQUE, COUDIERES, GENOUILLERES, PROTEGE-POIGNETS, GANTS). A UTILISER SUR LES VOIES PRIVEES HORS DE TOUTE CIRCULATION AUTOMOBILE. A UTILISER SUR DES SURFACES PLANES PROPRES, SECHES ET SANS OBSTACLES ET ELOIGNEES DES USAGERS DE LA ROUTE: POUR LES DEBUTANTS: SE FAIRE AIDER PAR UNE PERSONNE ADULTE. AVANT UTILISATION: VERIFIER LE REGLAGE DE L'ARTICLE ET QUE TOUS LES ELEMENTS DE CONNEXION SOIENT FIXES SOLIDEMENT. AVERTISSEMENT: NE PAS SAUTER DES OBSTACLES AVEC L'ARTICLE. NE PAS SUPPRIMER LE SYSTEME DE FREINAGE. NE PAS UTILISER DE NUIT. UTILISER AVEC LES DEUX MAINS SUR LE GUIDON. NE PAS UTILISER PIEDS NUS. INSTRUCTIONS D'ENTRETIEN ET DE STOCKAGE: AVANT CHAQUE UTILISATION, CONTROLER LE SERRAGE DE TOUTES LES VIS; DE TOUS LES ECROUS, ET DE TOUTES LEX FIXATIONS DES ROUES OU DU GUIDON. REMPLACER LES PIECES D'USURE (ROUES, FREINS, ETC...) PAR DES PIECES D'ORIGINE. GRAISSER REGULIEREMENT LES ROULEMENTS DES ROUS ET DES ARTICULATIONS. ELIMINER TOUT ARETE VIVE OCCASIONEE PENDANT L'UTILISATION: LAVER A L'EAU ET AU SAVON. NE PAS UTILISER DE SOLVANTS. NE PAS RANGER HUMIDE. (Note: there is a spelling error in the 4th sentence from the end where "roues" is spelled "rous". ) NOTE TO BE PRESERVED MINIMAL LOAD OF USE. 20 KG. CAUTION: TO USE WITH PROTECTION EQUIPEMENTS (HELMETS, COUDIERES, KNUCKLES, PROTEGE Wrists, GLOVES). TO USE ON WAYS PRIVEES OUT OF ANY MOTOR VEHICLE TRAFFIC. TO USE ON CLEAN PLANE SURFACES, CIGS AND WITHOUT OBSTACLES AND ELOIGNEES OF THE ROAD USERS. FOR THE DEBUTANTS: SE TO MAKE HELP BY AN ADULT PERSON. BEFORE USE: VERIFIER The REGLAGE OF The ARTICLE AND THAT ALL The ELEMENTS OF CONNECTION ARE FIXED FIRMLY. WARNING: NOT TO JUMP OF The OBSTACLES WITH THE ARTICLE. NOT TO REMOVE THE BRAKE. NOT TO USE OF NIGHT. TO USE WITH THE TWO HANDS ON THE HANDLEBAR. NOT TO USE NAKED FEET. INSTRUCTIONS OF MAINTENANCE AND STORAGE BEFORE THE EACH USE, TO CONTROL THE TIGHTENING OF ALL THE SCREWS, ALL THE NUTS, AND ALL THE FIXINGS OF THE WHEELS OR HANDLEBAR. TO REPLACE THE WEARING PARTS (WHEELS, BRAKES, ETC.) BY ORIGINAL PARTS. TO REGULARLY LUBRICATE THE BEARINGS OF THE ROUS AND THE ARTICULATIONS. TO ELIMINATE ANY ARRIS OCCASIONEE DURING The USE. TO WASH A WATER AND WITH THE SOAP. NOT TO USE SOLVENTS. NOT TO ARRANGE WET. --- end ---- From roisin_saunier@hotmail.com Tue Apr 8 11:42:17 2003 From: roisin_saunier@hotmail.com (Roisin Saunier) Date: Tue, 08 Apr 2003 12:42:17 +0200 Subject: [MT-List] Research into usability and MT Message-ID: Hello all, Can anyone point me to research papers covering the needs of (potential) MT users other than professional translators? All I have found to date are several papers where the authors recognize the need for research in this area. Many thanks, Róisín _________________________________________________________________ STOP MORE SPAM with the new MSN 8 and get 2 months FREE* http://join.msn.com/?page=features/junkmail From jeff.allen@free.fr Wed Apr 9 00:50:48 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Wed, 9 Apr 2003 01:50:48 +0200 Subject: [MT-List] Re: Research into usability and MT References: Message-ID: <001b01c2fe82$970843a0$a89f933e@home> > De : "Roisin Saunier" > Envoyé : Tuesday, April 08, 2003 12:42 PM > Objet : [MT-List] Research into usability and MT > Can anyone point me to research papers covering the needs of (potential) >MT users other than professional translators? All I have found to date are > several papers where the authors recognize the need for research in this > area. Róisín, Examples that you are looking for are in: MORLAND, Verne. Nutzlos, Bien Pratique or Muy Util? Business Users Speak Out on the Value of Pure MT. ASLIB 2002 conference. (this describes the use of MT by e-learning users through the NCR corporate university) ELLMAN, Jeremy et al. 1999. MABLe: a Multi-lingual Authoring Tool for Business Letters. Presented at the 21st Conference of "Translating and the Computer", sponsored by ASLIB, 10-11 November 1999, London. BROWN, Ralf (1998) Human-computer interactivity for users of the DIPLOMAT speech-to-speech system interface. Workshop on Embedded MT Systems, AMTA 98 conference. 28 October 1998, Langhorne, Pennsylvania. Available at http://www-2.cs.cmu.edu/~ralf/papers/amta98.ps.gz (this describes the use of MT and usability issues for military and peace-keeping operations users) HOGAN, Christopher. 1998. Corrected OCRed Text for MT. In proceedings of Workshop on Embedded MT Systems, AMTA 98 conference. 28 October 1998, Langhorne, Pennsylvania. (this describes the use of OCR combined with MT for information scanning/filtering needs) HOGAN, Christopher. 1999. "OCR for Minority Languages". In Proceedings of the 1999 Symposium on Document Image Understanding Technology, Annapolis, Maryland, April 1999, pp. 235-244. http://www-2.cs.cmu.edu/~chogan/sdiut99.zip (this describes the use of OCR combined with MT for information scanning/filtering needs) JOHNSON, Eduard. LinguaNet (TM) - Controlling Police Communication. 1996. In proceedings of the First International Controlled Language Applications Workshop (CLAW96), Louvain, Belgium. HANSEN, Inge G., & Henrik S. SORENSEN. 1998. LinguaNet - Embedded Translation Facilities in Cross-border Police Messaging. In proceedings of Workshop on Embedded MT Systems, AMTA 98 conference. 28 October 1998, Langhorne, Pennsylvania. http://www.cbs.dk/projects/linguanet/ SELIGMAN, Mark, Mary FLANAGAN, and Sophie TOOLE. 1998. Dictated Input for Broad-coverage Speech Translation. In proceedings of Workshop on Embedded MT Systems, AMTA 98 conference. 28 October 1998, Langhorne, Pennsylvania. (this describes speech recognition and speech synthesis combined with online chat translation) Look for other articles available concerning the Compuserve online chat translation work done by Mary FLANAGAN ZAJAC, Rémi & Mohamed BOUDEMAGH. Evaluation of MT of Financial Documents. Workshop: Using Evaluation within HLT programmes: Results and Trends. In proceedings of the Satellite Workshop of the Second International Conference on Language Resources and Evaluation (LREC2000), 31 May - 2 June 2000, Athens, Greece. pp 30-33. (order proceedings via ELDA www.elda.fr) Decrozant, Lisa and Clare Voss. (1998) Cross-linguistic Resources for MT Evaluation and Language Training. Paper presented at the International Conference on Natural Language Processing and Industrial Applications (NLP+IA 98), Moncton, New-Brunswick, Canada. 18-21 August 1998. Decrozant, Lisa and Clare Voss. 1999. Building a 'Tri-Text': Steps in the Conversion of a Hard Copy Document to an On-line Resource. In ELRA Newsletter. Vol 4 issue 1; January 1999, Paris: European Language Resources Association. pp. 10-11. (contact ELRA/ELDA at www.elda.fr) Frederking, Robert, Alexander Rudnicky, and Christopher Hogan. "Interactive Speech Translation in the DIPLOMAT Project," In Proceedings of the Spoken Language Translation Workshop at Association for Computational Linguistics (ACL 97), Madrid, Spain. pp. 61-66, 1997. VOSS, Clare & Carol VAN ESS-DYKEMA. 2000. When is an Embedded MT System "Good Enough" for Filtering? In proceedings of 2nd Workshop on Embedded MT Systems, 4 May 2000, held in conjunction with the ANLP/NAACL2000 Language Technology Joint Conference, Seattle, Washington. REEDER, Florence. 2000. At Your Service: Embedded MT as a Service. In proceedings of 2nd Workshop on Embedded MT Systems, 4 May 2000, held in conjunction with the ANLP/NAACL2000 Language Technology Joint Conference, Seattle, Washington. MENG, Helen et al. 2000. Mandarin-English Information (MEI): Investigating Translingual Speech Retrieval. In proceedings of 2nd Workshop on Embedded MT Systems, 4 May 2000, held in conjunction with the ANLP/NAACL2000 Language Technology Joint Conference, Seattle, Washington. TURCATO, Davide et al. 2000. Pre-processing Closed Captions for MT. In proceedings of 2nd Workshop on Embedded MT Systems, 4 May 2000, held in conjunction with the ANLP/NAACL2000 Language Technology Joint Conference, Seattle, Washington. Has anyone conducted studies on the use of portable translation units (like Franklin bilingual dictionary translators) and Pocket PC/Palm PDA MT software by tourists? Has anyone conducted studies on the use of translingual reservation systems (hotel reservations, flight reservations)? Regards, Jeff --------------------------------- Jeff ALLEN Paris, FRANCE Mobile Tel: (+33) 6 86 06 87 59 e-mail: jeff.allen@free.fr From WJHutchins@compuserve.com Wed Apr 9 12:12:06 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Wed, 9 Apr 2003 07:12:06 -0400 Subject: [MT-List] conference on Controlled Language Translation Message-ID: <200304090712_MC3-1-3394-B05@compuserve.com> To all with interest in MT and controlled language: The EAMT/CLAW 2003 conference on Controlled Language = Translation (Dublin 15-17 May, 2003) is still open for registration at www.http://www.eamt.org/eamt-claw03/. = Visit the conference web site to see the interesting and wide = ranging programme, the social events for participants and = to get information on accommodation etc. Do not miss it! John Hutchins 9 April From gerbl@pacbell.net Fri Apr 11 05:01:06 2003 From: gerbl@pacbell.net (Laurie Gerber) Date: Thu, 10 Apr 2003 21:01:06 -0700 Subject: [MT-List] Call for Exhibitors - MT Summit IX Sept 24-26, 2003, New Orleans Message-ID: <3E963E02.2040207@pacbell.net> *******************Call for Exhibitors******************** MT Summit IX New Orleans, LA, USA, September 23-27, 2003 http://www.mt-summit.org The Machine Translation Summit, hosted by the International Association for Machine Translation, is an exciting event for anyone interested in translation technology. This biennial event comes to the US only once every 6 years, and provides a forum for researchers, developers, vendors, users, shoppers, funders and policymakers from the U.S. and abroad to learn about new developments and best practices in each area of machine translation. The MT Summit attracts participants from Asia and Europe as well as the United States. The last couple of years have brought some exciting developments in many areas: evolutionary advances in MT research, lots of forward-thinking deployments of MT, significant improvements in existing commercial products, and a whole cohort of startup companies commercializing newer approaches to the core problems of machine translation. The organizing team for the exhibits at the 9th international MT Summit is energetically working on attracting more users and shoppers to this event with targeted local, national and international publicity. The key to increasingly lively and interesting commercial events depends on bringing together the people who need the technology and the vendors who can help them. When we polled users, potential users, and funders about what would make them want to attend the conference, it was primarily a good variety of vendors to talk to. In addition to MT developers and vendors, we are particularly eager to include • consultants and integrators who can help potential users to understand and make the best use of the technology that is available • providers of companion technologies that can leverage or provide leverage to machine translation, such as content management, globalization management systems, translation memory, and others. • Providers of component technologies that may be of interest to other developers and researchers, such as morphological analysers and generators, parsers, and others. **Exhibition dates** From: Tuesday, September 23 (opening at 6:00 p.m. - conference reception will be in and around exhibits.) To: Friday, September 26 (closing at 5:00 pm when the conference concludes.) Exhibition rooms available for setup: Tuesday, September 23, 12:00 noon. **Securing and Paying for Your Exhibit Space** The exhibit fee is $600 for Corporate and Institutional members of any IAMT regional division (AAMT, AMTA, EAMT), $800 for non-members, and includes: - Tabletop exhibit area - Sign for each exhibitor - Skirted table and two chairs Monitors and other display equipment can be rented from Swank Audio Visuals. (504) 529-4797 MT Summit IX in Brief The 9th International Association for Machine Translation Summit conference will be held at the elegant and historic Fairmont hotel in New Orleans, Louisiana, September 23-27, 2003. Expected attendance at the conference is between 200 and 300 people, including MT shoppers, representatives of the U.S. Government, consultants, localization companies and tool providers, MT researchers and developers. Up to two representatives of each exhibitor may staff the exhibit, and are welcome to attend the conference sessions. If an exhibitor wishes to have more than two representatives, they are asked to register as regular conference participants. Exhibitors may also register for the banquet, tutorials, and workshops. Further information, and a registration form, are available at: http://www.mt-summit.org Please send questions to: Laurie Gerber or Keith Miller, Exhibits Coordinators, MT Summit IX Laurie Gerber Tel: 619-200-8344 Email: gerbl@pacbell.net Keith Miller Email : Keith@mitre.org From gerbl@pacbell.net Fri Apr 11 07:00:42 2003 From: gerbl@pacbell.net (Laurie Gerber) Date: Thu, 10 Apr 2003 23:00:42 -0700 Subject: [MT-List] Call for Tutorials at MT Summit IX - Sept 23, 2003 Message-ID: <3E965A0A.8060005@pacbell.net> **************Call for Tutorial Proposals********************** Machine Translation Summit IX New Orleans, LA, USA, September 23-28, 2003 http://www.mt-summit.org Tutorials at the MT Summit provide an opportunity for the experts to share their insights in sessions with small groups of interested participants. Tutorials will be held Tuesday, September 23, the day before the main conference begins. We plan an exciting tutorial program at MT Summit IX, with tutorials focused on the interests of end users, decision makers, integrators, and MT developers alike. We envision a tutorial program consisting of " one introductory tutorial for people new to the field, in the spirit of Introduction to machine translation and translation technology " two technically-focused tutorials on MT development / enhancement " four user-focused tutorials in areas such as o Latest developments in commercially available MT o Evaluation and selection of MT products / assessment of individual tools for specific business needs We already have three world-class tutorials lined up on the topics of translation project management, using MT in practice, and writing for translation, and we are currently working to round out the tutorial program for the Summit. To this end, proposals are now being solicited! Submissions should be made to the Tutorials Co-Chairs, Laurie Gerber gerbl@pacbell.net and Keith Miller keith@mitre.org by April 30th. They should include the proposed title of the tutorial, a brief outline of the contents, the name(s) of the tutorialist(s) and CV(s) or other account of qualifications. Important Dates April 30, 2003 Deadline for Tutorial proposals May 15, 2003 Notification of acceptance of proposal September 23, 2003 MT Summit Tutorials September 24-26, 2003 Main MT Summit conference sessions September 27 2003 MT Summit post-conference workshops For general conference information and further details as it becomes available, visit the MT Summit web site http://www.mt-summit.org From Andy.Way" The EAMT/CLAW 2003 conference on Controlled Translation (Dublin 15-17 May, 2003) is still open for registration at http://www.eamt.org/eamt-claw03/. Visit the conference web site to see the programme, social events and to get information on accommodation etc. Andy Way From harold.somers@umist.ac.uk Fri Apr 18 09:55:04 2003 From: harold.somers@umist.ac.uk (Harold Somers) Date: Fri, 18 Apr 2003 09:55:04 +0100 Subject: [MT-List] Survey of online MT Message-ID: There is an online questionnaire about the use of online MT which a final-year student of mine has set up for her project. If you can spare a few minutes, please fill it in (and also ask any of your friends who use online MT to do so too). Many thanks. The survey is at http://www.ccl.umist.ac.uk/onlinemt.html (It's really aimed at naive users, but because of delays in getting the page online, I think Sarah will be happy to receive replies from anyone! The project has to be finished in a few weeks time, so please HURRY!) ======================================== Harold Somers Professor of Language Engineering, UMIST From bermanv@zahav.net.il Sat Apr 19 07:47:20 2003 From: bermanv@zahav.net.il (Vadim Berman) Date: Sat, 19 Apr 2003 08:47:20 +0200 Subject: [MT-List] New website dedicated to machine translation is up Message-ID: <001201c3063f$87db2c20$3fc608d5@bermanv> No Limits, a new website dedicated to the subject of machine translation, is up. History, analysis, speculations on the future of MT, tons of usable and off-the-beaten-track resources which are constantly updated. The author emphasizes on the commercial aspect of machine translation. No Limits is hosted by http://AI-Depot.com at http://nolimits.AI-Depot.com . Best regards, Vadim Berman From jeff.allen@free.fr Sat Apr 19 16:02:07 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Sat, 19 Apr 2003 17:02:07 +0200 Subject: [MT-List] another MT software review article References: <00da01c2f8df$d1e47140$0100007f@home> Message-ID: <07a101c3081b$bbf76e00$df9c933e@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_06FE_01C30695.6902F0C0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear all, Another software review appeared in Language International (magazine is = unfortunately now discontinued since Dec 2002). MOISL, Angela. 2002. A Fresh Look at MT: a professional translator = reviews the Personal Translator MT application. In Language = International, Vol. 14, No. 5, October 2002. Regards, Jeff --------------------------------- Jeff ALLEN Paris, France Mobile Tel: (+33) 6 86 06 87 59 e-mail: jeff.allen@free.fr =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Wednesday, April 02, 2003 7:47 AM Objet : [MT-List] more articles in MLCT on MT and translation = technologies =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Friday, March 14, 2003 10:40 PM Objet : [MT-List] updated links: articles in MLCT on MT, TM, = postediting ------=_NextPart_000_06FE_01C30695.6902F0C0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Dear all,
Another software review appeared in Language = International=20 (magazine is unfortunately now discontinued since Dec = 2002).
 
MOISL, Angela. 2002. A Fresh Look at MT: a = professional=20 translator reviews the Personal Translator MT application. In Language=20 International, Vol. 14, No. 5, October 2002.
 
Regards,
Jeff

---------------------------------
Jeff = ALLEN
Paris,=20 France
Mobile Tel: (+33) 6 86 06 87 59
e-mail: jeff.allen@free.fr
 
----- Message d'origine -----
De=20 : Jeff=20 Allen
Envoy=E9 : Wednesday, = April 02, 2003=20 7:47 AM
Objet : [MT-List] more articles = in MLCT=20 on MT and translation technologies

<snip>
 
  ----- = Message=20 d'origine -----
De : Jeff = Allen=20
Envoy=E9 : Friday, March 14, 2003 10:40 PM
Objet : [MT-List] updated links: articles in MLCT on MT, = TM,=20 postediting

<snip>
------=_NextPart_000_06FE_01C30695.6902F0C0-- From toby@ccl.umist.ac.uk Tue Apr 22 12:20:33 2003 From: toby@ccl.umist.ac.uk (Toby Peers) Date: Tue, 22 Apr 2003 12:20:33 +0100 Subject: [MT-list] Is this hidden text technique original? Message-ID: <1051010433.3ea5258175dd6@webmail1.umist.ac.uk> Dear All, I've set up a web page which includes invisible text, which helps force the translation of the word open. It does this by including an invisible sentence around the word. You can see this effect by translating the page below in the google language tools page. http://mull.ccl.umist.ac.uk/staff/toby/Partial_Translation_in_CALL/testing_systr an/invisible_text.html there are two instances of 'open mail', which appear the same on the untranslated page but are both translated differently due to the invisible sentences around them. Is this technique original, useful, pointless, etc ?? The html is below Test of invisible words in MT The open door was red. The mail was light.

open mail

The open door was red. The mail was light. Thanks Toby ************************************************************* Toby Peers Centre for Computational Linguistics UMIST PO Box 88 Tel: +44 161 200 3077 Manchester M60 1QD Fax: +44 161 200 3091 UK toby@ccl.umist.ac.uk ************************************************************* From nadamides@aslib.com" TRANSLATING AND THE COMPUTER 25 CONFERENCE, 20-21 November 2003, London This conference is one of the few international events which focuses on the user aspects of translation software and as such has been particularly beneficial to a very wide audience including translators, business managers, researchers and language experts. The event, which is organised by Aslib, has the support of EAMT, IAMT, BCS - Natural Language Group, Institute of Linguists. Once again, this year the conference will address the latest developments in translation (and translation-related) software. It will address the needs of the following conference attendees: industry public administration agencies freelancers development This call for papers invites abstracts of papers to be presented at the conference. The papers (and the presentations) should focus on the user aspects of translation or translation-related software rather than on theoretical issues. Presentations accompanied by demonstrations are espe cially welcome. TOPICS The range of topics includes (but is not limited to) use of MT systems machine-aided translation and translation aids controlled languages and their use in MT speech translation terminology localisation multilingual document management/workflow case studies of technology-based solutions the Internet and translation aids/services the value of "free" versus "charging" services/sites on the Internet SUBMISSION GUIDELINES Authors are required to submit an abstract of a MINIMUM of 500 words of the paper they would like to present, together with an outline of the structure of the paper and short BIOGRAPHY. Abstracts should be sent by POST or EMAIL before 15th June 2003 to: Nicole Adamides, Conference Organiser Aslib, The Association for Information Management Temple Chambers, 3-7 Temple Avenue, London, EC4Y 0HP Tel: +44(0) 20 7583 8900 Fax: +44 (0) 20 7583 8401 Email: nadmides@aslib.com The abstracts will be considered by the Programme Chairs, namely: Daniel Grasmick and Chris Pyne, SAP; Professor Ruslan Mitkov, University of Wolverhampton and Olaf-Michael Stefanov, United Nations. The authors of abstracts will be notified of acceptance or rejection of their submissions by 1 August 2003. The full length versions of the accepted papers (authors will be provided with detailed camera-ready copy guidelines) will be included in the conference proceedings and must be submitted by 10th October 2003. From jyang@systransoft.com Fri Apr 25 01:59:00 2003 From: jyang@systransoft.com (Jin Yang) Date: Thu, 24 Apr 2003 17:59:00 -0700 Subject: [MT-List] MT Summit IX: 4th CFP Message-ID: Hi, Could you please post the following? Thanks. webmaster@amtaweb.org MACHINE TRANSLATION SUMMIT IX September 23-27, 2003 New Orleans, USA http://www.mt-summit.org CALL FOR PAPERS The ninth Machine Translation Summit, organized by the International Association for Machine Translation (IAMT) and hosted by the Association for Machine Translation in the Americas (AMTA), will be held in New Orleans, Louisiana, from 23 to 27 September 2003. MT Summit IX hereby invites original submissions on all aspects machine and machine-aided translation. Please consult the conference Web site for the complete Call for Papers, including details on the different categories of submissions. Important Reminder: the deadline for the submission of papers is ***May 11, 2003***. From bermanv@zahav.net.il Thu Apr 24 20:04:55 2003 From: bermanv@zahav.net.il (Vadim Berman) Date: Thu, 24 Apr 2003 21:04:55 +0200 Subject: [MT-list] Is this hidden text technique original? References: <1051010433.3ea5258175dd6@webmail1.umist.ac.uk> Message-ID: <000101c30b1e$26e6f320$0f1bfea9@vb> Toby, What effect do you get? Do you mean, that your purpose is translating a particular word-sense of "open"? Maybe the software you've used related to the hidden text as several chunks of text? Best regards, Vadim Berman ----- Original Message ----- From: Toby Peers To: Sent: Tuesday, April 22, 2003 1:20 PM Subject: [MT-list] Is this hidden text technique original? > Dear All, > > I've set up a web page which includes invisible text, which helps force the > translation of the word open. It does this by including an invisible sentence > around the word. You can see this effect by translating the page below in the > google language tools page. > > http://mull.ccl.umist.ac.uk/staff/toby/Partial_Translation_in_CALL/testing_s ystr > an/invisible_text.html > > there are two instances of 'open mail', which appear the same on the > untranslated page but are both translated differently due to the invisible > sentences around them. Is this technique original, useful, pointless, etc ?? > > The html is below > > > > > Test of invisible words in MT > > > > The open door > was red. > The mail was light. >
>

> open mail >

>

> The open door was red. The mail was light. > > > > Thanks > Toby > > ************************************************************* > Toby Peers > Centre for Computational Linguistics > UMIST > PO Box 88 Tel: +44 161 200 3077 > Manchester M60 1QD Fax: +44 161 200 3091 > UK toby@ccl.umist.ac.uk > > ************************************************************* > > -- > For MT-List info, see http://www.eamt.org/mt-list.html From stopping@bizwonk.com Fri Apr 25 20:16:19 2003 From: stopping@bizwonk.com (Suzanne M. Topping) Date: Fri, 25 Apr 2003 15:16:19 -0400 Subject: [MT-list] Is this hidden text technique original? Message-ID: <427F53DA8F48E9498ADF0F868763F88C1D462E@wonkserver1.bizwonk.com> Is the hidden text attempting to provide additional context to allow for = higher accuracy in the translation of the viewable text? > -----Original Message----- > From: Vadim Berman [mailto:bermanv@zahav.net.il]=20 > Sent: Thursday, April 24, 2003 3:05 PM > To: mt-list@eamt.org > Subject: Re: [MT-list] Is this hidden text technique original? >=20 >=20 >=20 > Toby, >=20 > What effect do you get? Do you mean, that your purpose is=20 > translating a particular word-sense of "open"? >=20 > Maybe the software you've used related to the hidden text as=20 > several chunks of text? >=20 > Best regards, > Vadim Berman >=20 > ----- Original Message ----- > From: Toby Peers > To: > Sent: Tuesday, April 22, 2003 1:20 PM > Subject: [MT-list] Is this hidden text technique original? >=20 >=20 > > Dear All, > > > > I've set up a web page which includes invisible text, which helps=20 > > force > the > > translation of the word open. It does this by including an invisible > sentence > > around the word. You can see this effect by translating the=20 > page below=20 > > in > the > > google language tools page. > > > > > http://mull.ccl.umist.ac.uk/staff/toby/Partial_Translation_in_ > CALL/testing_s > ystr > > an/invisible_text.html > > > > there are two instances of 'open mail', which appear the=20 > same on the=20 > > untranslated page but are both translated differently due to the=20 > > invisible sentences around them. Is this technique=20 > original, useful,=20 > > pointless, etc > ?? > > > > The html is below > > > > > > > > > > Test of invisible words in MT > > > > > > > > The open > style=3D"display:none"> > door > > was red. > > The mail was=20 > light.
> >

> > open mail > >

> >

> > The open door was red. The mail was light. > > > > > > > > Thanks > > Toby > > > > ************************************************************* > > Toby Peers > > Centre for Computational Linguistics > > UMIST > > PO Box 88 Tel: +44 161 200 3077 > > Manchester M60 1QD Fax: +44 161 200 3091 > > UK toby@ccl.umist.ac.uk > > > > ************************************************************* > > > > -- > > For MT-List info, see http://www.eamt.org/mt-list.html >=20 >=20 > --=20 > For MT-List info, see http://www.eamt.org/mt-list.html >=20 From toby@ccl.umist.ac.uk Sat Apr 26 10:48:23 2003 From: toby@ccl.umist.ac.uk (Toby Peers) Date: Sat, 26 Apr 2003 10:48:23 +0100 Subject: [MT-list] Is this hidden text technique original? In-Reply-To: <427F53DA8F48E9498ADF0F868763F88C1D462E@wonkserver1.bizwonk.com> References: <427F53DA8F48E9498ADF0F868763F88C1D462E@wonkserver1.bizwonk.com> Message-ID: <1051350503.3eaa55e7d999f@webmail1.umist.ac.uk> Suzanne > Is the hidden text attempting to provide additional context to allow for > higher accuracy in the translation of the viewable text? Yes, and it seems to work. It might be possible to try this with the HTML comment tags, i.e open etc .... but I haven't tried this. The only problems with these methods are that some web developers have used it to embed high frequency search terms in their pages. This means you may type 'flights' into lycos and end up getting something completely irrelevant popping up high in your list of hits. I discovered this while searching for '"invisible text" HTML' on GOOGLE. I was looking for ways to embed invisible text. It was in the groups where there was quite alot of discussion of this 'evil' way to fool the search engines. If I recall correctly (though I may be wrong) they stated that many search engines have become wise to it now and have stopped listing sites with invisible text. This might be a bit of a killer for the whole idea. Toby > > > -----Original Message----- > > From: Vadim Berman [mailto:bermanv@zahav.net.il] > > Sent: Thursday, April 24, 2003 3:05 PM > > To: mt-list@eamt.org > > Subject: Re: [MT-list] Is this hidden text technique original? > > > > > > > > Toby, > > > > What effect do you get? Do you mean, that your purpose is > > translating a particular word-sense of "open"? > > > > Maybe the software you've used related to the hidden text as > > several chunks of text? > > > > Best regards, > > Vadim Berman > > > > ----- Original Message ----- > > From: Toby Peers > > To: > > Sent: Tuesday, April 22, 2003 1:20 PM > > Subject: [MT-list] Is this hidden text technique original? > > > > > > > Dear All, > > > > > > I've set up a web page which includes invisible text, which helps > > > force > > the > > > translation of the word open. It does this by including an invisible > > sentence > > > around the word. You can see this effect by translating the > > page below > > > in > > the > > > google language tools page. > > > > > > > > http://mull.ccl.umist.ac.uk/staff/toby/Partial_Translation_in_ > > CALL/testing_s > > ystr > > > an/invisible_text.html > > > > > > there are two instances of 'open mail', which appear the > > same on the > > > untranslated page but are both translated differently due to the > > > invisible sentences around them. Is this technique > > original, useful, > > > pointless, etc > > ?? > > > > > > The html is below > > > > > > > > > > > > > > > Test of invisible words in MT > > > > > > > > > > > > The open > > style="display:none"> > > door > > > was red. > > > The mail was > > light.
> > >

> > > open mail > > >

> > >

> > > The open door was red. The mail was light. > > > > > > > > > > > > Thanks > > > Toby > > > > > > ************************************************************* > > > Toby Peers > > > Centre for Computational Linguistics > > > UMIST > > > PO Box 88 Tel: +44 161 200 3077 > > > Manchester M60 1QD Fax: +44 161 200 3091 > > > UK toby@ccl.umist.ac.uk > > > > > > ************************************************************* > > > > > > -- > > > For MT-List info, see http://www.eamt.org/mt-list.html > > > > > > -- > > For MT-List info, see http://www.eamt.org/mt-list.html > > > > -- > For MT-List info, see http://www.eamt.org/mt-list.html > ************************************************************* Toby Peers Centre for Computational Linguistics UMIST PO Box 88 Tel: +44 161 200 3077 Manchester M60 1QD Fax: +44 161 200 3091 UK toby@ccl.umist.ac.uk ************************************************************* From gerbl@pacbell.net Sun Apr 27 04:49:25 2003 From: gerbl@pacbell.net (Laurie Gerber) Date: Sat, 26 Apr 2003 20:49:25 -0700 Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. Message-ID: <3EAB5345.2020509@pacbell.net> Dear MT-ers, Are there any iron-clad, authoritative definitions of Statistical MT, Example-based MT and/or Data-driven MT that would provide sufficient and necessary conditions for MT systems to claim membership in any of the above categories? How about working definitions? I'd even be happy with a different set of categories with definitions, if I could use them for real MT systems! In putting together an article on the emerging data-driven MT systems for MTNI, I found that very few of the emerging or updated MT systems fit my previous notions of what such systems would be. For example: 1) Many systems claim a statistical component. What does the MT community think is necessary to make such a claim? What is necessary for a system to claim to be "statistical MT"? 2) There are a number of developers who describe their systems as example-based, but all of the examples or patterns are hand-built from linguist observations of data with apparently no automated learning component. In addition, patterns may be abstracted to phrase-structure rules. At some point, it starts to look a lot like rule-based MT, although it may be more like direct MT than transfer. 3) Any thoughts on definitions for "hybrid" systems? Thanks!!! Laurie Gerber Editor Machine Translation News International From carl@iai.uni-sb.de Sun Apr 27 18:44:28 2003 From: carl@iai.uni-sb.de (Michael Carl) Date: Sun, 27 Apr 2003 19:44:28 +0200 Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. References: <3EAB5345.2020509@pacbell.net> Message-ID: <3EAC16FC.5050904@iai.uni-sb.de> --------------060502020302010504090106 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Hi Laurie, in my understanding, EBMT and SMT are both different instantiations of Data-driven MT: while SMT (actually: probabilistic MT) systems are rooted in the IBM models, EBMT is based on - or at least can be better explained by - analogical reasoning. 1) The fact that an MT system uses a statistical component does not make it a statistical one, in the same way a system does not become rule-based if it uses a (set of) rule(s). 2) There is a paper from Davide Turcato & Fred Popowich: What is Example-Based Machine Translation? http://www.eamt.org/summitVIII/workshop-papers.html making a very similar point to what you say. 3) hybrid systems integrate different (computational) paradigms where the author(s) want to stress that these components are equally important. best regards, Michael Laurie Gerber wrote: > Dear MT-ers, > > Are there any iron-clad, authoritative definitions of > Statistical MT, Example-based MT and/or Data-driven MT > that would provide sufficient and necessary conditions for > MT systems to claim membership in any of the above > categories? How about working definitions? I'd even > be happy with a different set of categories with definitions, > if I could use them for real MT systems! > > In putting together an article on the emerging data-driven > MT systems for MTNI, I found that very few of the emerging or > updated MT systems fit my previous notions of what such > systems would be. For example: > 1) Many systems claim a statistical component. What does the > MT community think is necessary to make such a claim? What > is necessary for a system to claim to be "statistical MT"? > 2) There are a number of developers who describe their systems > as example-based, but all of the examples or patterns are hand-built > from linguist observations of data with apparently no automated learning > component. In addition, patterns may be abstracted to phrase-structure > rules. At some point, it starts to look a lot like rule-based MT, > although it may be more like direct MT than transfer. > 3) Any thoughts on definitions for "hybrid" systems? > > Thanks!!! > > Laurie Gerber > Editor > Machine Translation News International > > > --------------060502020302010504090106 Content-Type: text/html; charset=us-ascii Content-Transfer-Encoding: 7bit Hi Laurie,

in my understanding, EBMT and SMT are both different instantiations
of Data-driven MT: while SMT (actually: probabilistic MT) systems are
rooted in the IBM models, EBMT is based on - or at least can be better
explained by - analogical reasoning.
1) The fact that an MT  system uses a  statistical component does not make
it a statistical one, in the same way a system does not become rule-based
if it uses a (set of) rule(s).
2) There is a paper from Davide Turcato & Fred Popowich:
What is Example-Based Machine Translation?
http://www.eamt.org/summitVIII/workshop-papers.html
making a very similar point to what you  say.
3) hybrid  systems integrate different (computational) paradigms where the
author(s) want to stress that these components are equally important.

best regards,
Michael



Laurie Gerber wrote:

Dear MT-ers,

Are there any iron-clad, authoritative definitions of
Statistical MT, Example-based MT and/or Data-driven MT
that would provide sufficient and necessary conditions for
MT systems to claim membership in any of the above
categories? How about working definitions? I'd even
be happy with a different set of categories with definitions,
if I could use them for real MT systems!

In putting together an article on the emerging data-driven
MT systems for MTNI, I found that very few of the emerging or
updated MT systems fit my previous notions of what such
systems would be. For example:
1) Many systems claim a statistical component. What does the
MT community think is necessary to make such a claim?  What
is necessary for a system to claim to be "statistical MT"?
2) There are a number of developers who describe their systems
as example-based, but all of the examples or patterns are hand-built
from linguist observations of data with apparently no automated learning
component. In addition, patterns may be abstracted to phrase-structure
rules. At some point, it starts to look a lot like rule-based MT,
although it may be more like direct MT than transfer.
3) Any thoughts on definitions for  "hybrid" systems?

Thanks!!!

Laurie Gerber
Editor
Machine Translation News International




--------------060502020302010504090106-- From ref@cs.cmu.edu Mon Apr 28 15:32:21 2003 From: ref@cs.cmu.edu (Robert Frederking) Date: Mon, 28 Apr 2003 10:32:21 -0400 Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. In-Reply-To: Your message of Sun, 27 Apr 2003 19:44:28 +0200. <3EAC16FC.5050904@iai.uni-sb.de> Message-ID: <27987.1051540341@lti.cs.cmu.edu> Hi Laurie, When I teach my part of the MT course here, I use the notion of a space of MT systems. There are certain points in that space that are clearly "Example-based" or "Statistical" or "rule-based", but many (perhaps most) real systems fall into gray areas in between the obvious examples. So I think that in your discussion, it would make sense to talk about whether a system is "closer to" the pure EBMT or pure Statistical or pure Rule-based point in that space. That said, I also make some definitions: -- an EBMT system is one that uses the parallel corpus at run-time, as opposed to a model trained in advance from the corpus (whether or not it uses careful mathematical justifications) -- a Statistical system is one that uses careful statistical justifications (these have so far usually used a statistical model built during training, and *not* the corpus, at runtime, but this has started to change recently) -- a Rule-based system uses a set of discrete rules. (If they are built in advance, the distinction from EBMT is clear; if a system automatically builds rules from a corpus at runtime, the distinction from EBMT gets gray.) You can easily create hypothetical systems that straddle each of these boundaries, and in fact a number of the statistical systems in the current DARPA MT evals build a phrase-based component at runtime from the corpus, which I think makes them *both* SMT and EBMT at the same time. I hope that's helpful. Bob From ref@cs.cmu.edu Mon Apr 28 19:11:52 2003 From: ref@cs.cmu.edu (Robert Frederking) Date: Mon, 28 Apr 2003 14:11:52 -0400 Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. In-Reply-To: Your message of Mon, 28 Apr 2003 10:32:21 -0400. <27987.1051540341@lti.cs.cmu.edu> Message-ID: <2772.1051553512@lti.cs.cmu.edu> One other thing: in terms of examples, there are obvious examples in research systems, although perhaps not in commercial systems. The CMU EBMT system (Ralf Brown is the main contact here) is pretty much purely example-based, although it uses a statistical LM to select outputs. (And more recent versions include the ability to add hand-crafted generalization rules.) The IBM model 1-5 systems are (in their earlier forms) clearly statistical systems, although later versions included hand-built rule-based components. They both follow the pattern of becoming more hybrid over time. Bob From hajic@ufal.ms.mff.cuni.cz Mon Apr 28 20:17:00 2003 From: hajic@ufal.ms.mff.cuni.cz (Jan Hajic) Date: Mon, 28 Apr 2003 21:17:00 +0200 (CEST) Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. In-Reply-To: <2772.1051553512@lti.cs.cmu.edu> Message-ID: I can but support the first email from Robert about the system classification, but as one of the insiders I have to correct the following: On Mon, 28 Apr 2003, Robert Frederking wrote: ... > The IBM model 1-5 systems are (in their earlier forms) clearly > statistical systems, although later versions included hand-built > rule-based components. No. They were 'fundamentalistically' statistical all the way down to around 1995 when the project was stopped. All the numerous analysis and generation steps that were added later (for WSD, for example, or even for word selection between electric/electrical etc. in the generation phase) were all statistical in the very sense of the definition. > > They both follow the pattern of becoming more hybrid over time. Not the IBM one. -- Jan > > Bob > > > From ref@cs.cmu.edu Tue Apr 29 20:25:27 2003 From: ref@cs.cmu.edu (Robert Frederking) Date: Tue, 29 Apr 2003 15:25:27 -0400 Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. In-Reply-To: Your message of Mon, 28 Apr 2003 21:17:00 +0200. Message-ID: <24948.1051644327@lti.cs.cmu.edu> > > They both follow the pattern of becoming more hybrid over time. > > Not the IBM one. Well, that differs from what has been reported in the past. The morphology processing wasn't rule-based? Or the word reordering used in later versions? Bob From hajic@ufal.ms.mff.cuni.cz Tue Apr 29 20:59:42 2003 From: hajic@ufal.ms.mff.cuni.cz (Jan Hajic) Date: Tue, 29 Apr 2003 21:59:42 +0200 (CEST) Subject: [MT-List] Definitions of statistical, example-based, data-driven MT, etc. In-Reply-To: <24948.1051644327@lti.cs.cmu.edu> Message-ID: On Tue, 29 Apr 2003, Robert Frederking wrote: > > > They both follow the pattern of becoming more hybrid > > > over time. > > > > Not the IBM one. > > Well, that differs from what has been reported in the > past. The morphology processing wasn't rule-based? Or Well if you insist yes, French and English morphology was dictionary-based (no rules though!). But there is no major single system that works that would use really *statistical* morphology, or is it (apart form recent Yarowsky & al. work on low-density languages)? Moreover, Candide did use morphology almost from the beginning, or at least for all the DARPA competitions, and that has not changed in the course of time. (And as expected, tagging was always statistical there.) So using a MRD converted to a morphological dictionary (and yes, partially hand-checked) does not constitute, in my opinion, grounds for calling the system "hybrid". > the word reordering used in later versions? As far as I remember, no. Local word reordering (such as swapping French -de- construction when translating into) has been trained off data as well, and more global "reordering" was simply part of the distortion models. Jan > > Bob > > From hovy@ISI.EDU Thu May 1 03:22:48 2003 From: hovy@ISI.EDU (Eduard Hovy) Date: Wed, 30 Apr 2003 19:22:48 -0700 Subject: [MT-List] Defs for types of MT In-Reply-To: <20030429193108.CF5E553A35@pairlist.net> References: <20030429193108.CF5E553A35@pairlist.net> Message-ID: Hi all, Laurie gerber asks an interesting question: > Are there any iron-clad, authoritative definitions of > Statistical MT, Example-based MT and/or Data-driven MT > that would provide sufficient and necessary conditions for > MT systems to claim membership in any of the above > categories? How about working definitions? I'd even > be happy with a different set of categories with definitions, > if I could use them for real MT systems! Bob Frederking gives a nice answer: >-- an EBMT system is one that uses the parallel corpus at run-time, > as opposed to a model trained in advance from the corpus > (whether or not it uses careful mathematical justifications) >-- a Statistical system is one that uses careful statistical justifications > (these have so far usually used a statistical model built > during training, and *not* the corpus, at runtime, but this > has started to change recently) >-- a Rule-based system uses a set of discrete rules. (If they are > built in advance, the distinction from EBMT is clear; if a system > automatically builds rules from a corpus at runtime, the > distinction from EBMT gets gray.) I think Bob said it all. But it seemed interesting to me to wonder why one should focus not on the processing but rather on the core knowledge resource(s) used to perform the translation (as Bob indeed does). I think it's because it's the nature of the knowledge that determines what processing is most naturally used. This leads one to differentiate between the two stages of an MT system: building it and running it. All MT uses knowledge that specifies transformations of source into target (usually in a series of steps). Whether this knowledge is encoded as (traditional) rules or as EBMT-style patterns or as probabilistic tables, it seems increasingly the case that you can use either manual or automated (learning) methods to acquire the knowledge. This is true as much for IBM-style bilingual wordlists to EBMT-style regular expression phrase patterns (see Ravichandran and Hovy, ACL02 and follow-on this year) to Yamada-Knight-style tree transformations (see Yamada's thesis and Knight and Marcu, ACL01). Depending on how the knowledge is encoded, running the system then looks more 'statistical' or 'rule-based' , but that's a bad name for a basic difference in paradigm. What has been called statistical follows the paradigm of overgenerating alternatives and then ranking them, ending with a list of alternatives, each with a score. What has been called rule-based follows the paradigm of producing one output, and only on rejection allowing some backtracking to produce an alternative. Seen this way, you can see the two paradigms are not necessarily opposites; you can convert one to the other by appropriate (not always large) changes in parameterization and/or architecture. Bottom line: I think it's misleading to say there are fundamentally different systems. There is a basic method of acquiring the knowledge (manual vs learning; some of learning is probabilistic). There is the basic method of applying the knowledge. The core difference is in what the encoding of the knowledge looks like, and how it requires the processing to proceed most easily. But one form can slide into another. E -- ---------------------------------------------------------------------------- Eduard Hovy email: hovy@isi.edu USC Information Sciences Institute tel: 310-448-8731 4676 Admiralty Way fax: 310-823-6714 Marina del Rey, CA 90292-6695 project homepage: http://www.isi.edu/natural-language/nlp-at-isi.html From mlf@dlsi.ua.es Thu May 1 08:02:23 2003 From: mlf@dlsi.ua.es (Mikel L. Forcada) Date: 01 May 2003 09:02:23 +0200 Subject: [MT-List] CFP: Teaching Translation Technologies and Tools workshop at MT Summit IX Message-ID: <1051772553.3125.11.camel@rabassa> Teaching Translation Technologies and Tools (T4) A workshop at MT Summit IX CALL FOR PAPERS INTRODUCTION In September 2003, the workshop Teaching Translation Techniques and Tools will be held as part of the IX Machine Translation Summit in New Orleans, Louisiana. This is the third workshop addressing the teaching of machine translation; its predecessors were the Teaching Machine Translation workshop at the MT Summit VIII in Santiago de Compostela in September 2001 (http://www.dlsi.ua.es/tmt/) and, more recently, the 6th EAMT Workshop "Teaching Machine Translation" held in Manchester in November 2002 (http://www.ccl.umist.ac.uk/events/eamt-bcs/). This third workshop has an expanded scope which will not only address machine translation but also computer-aided translation technologies and tools. The workshop will provide an opportunity for machine translation instructors attending MT Summit VIII to exchange their experience by presenting papers or demonstrations describing the tools and techniques they use in the classroom or in the laboratory. SCHEDULE AND VENUE This one-day workshop will take place on September 27, after the regular conference sessions end. Please visit the Workshop website at http://www.dlsi.ua.es/t4/ for updates. PAPERS The Teaching Translation Technologies and Tools workshop seeks original papers in all aspects of the instruction of machine translation. Topics of interest include, but are not limited to: * why and to whom should MT be taught? * teaching the theoretical background of MT: linguistics, computer science, translation theory * addressing preconceptions about MT in the classroom * the use of commercial MT programs in hands-on teaching * teaching machine translation strategies to non-computational students * web-based distance learning of MT * teaching pre- and post-MT skills to MT users * teaching MT evaluation * building modules or `toy' MT systems in the laboratory * experiences on the evaluation of MT instruction * the role of MT in language learning * translation studies and MT * teaching MT-related techniques: translation memory, alignment tools, terminological databases, use of corpora, etc. The working language of the workshop will be English. Papers should describe research or experiences in any of the topics mentioned, should not be longer than 3000 words, and should nicely fit in 8 pages. In view of the main topics of the workhop, presentations may have a substantial demonstrative component. Electronic submissions (PDF files) are strongly preferred. As the reviewing process will be blind, authors are requested to keep their papers anonymous. This means that these submissions should NOT include the author's name; rather, papers should be identified only by their title. In addition to the electronic file containing their paper (including an abstract), authors must also submit a separate cover page with the following information: * paper title, * author(s)' name(s), affiliation(s), address(es), and e-mail address(es), * for demonstrative presentations: the hardware, software and network requirements for the demonstration. The two electronic files should be attached to an email and sent to mlf@dlsi.ua.es . We will acknowledge receipt of all papers received before the deadline and issue a submission number to each author. Please refer to that number in all subsequent correspondence. Anyone who is unable to make an electronic submission is asked to contact Mikel L. Forcada at mlf@dlsi.ua.es . Guidelines and style files for the preparation of the final camera-ready copy will be made available to the authors of accepted submissions in due time at the MT Summit IX site. If you cannot submit electronically, please mail 4 hardcopies of the paper to: Mikel L. Forcada, Teaching Translation Technologies and Tools at MT Summit IX, Dept. Llenguatges i Sistemes Inform=E0tics, Universitat d'Alacant, E-03071 Alacant, Spain IMPORTANT DATES Please make a note of these important dates: * Paper submission deadline: June 1, 2003 * Notification of acceptance: June 30, 2003 * Final camera-ready copy deadline: July 31, 2003 ORGANIZERS * Mikel L. Forcada , (mlf@dlsi.ua.es ), Departament de Llenguatges i Sistemes Inform=E0tics, Universitat d'Alacant, E-03071 Alacant, Spain. * Harold Somers , (Harold.Somers@umist.ac.uk ), Centre for Computational Linguistics, Department of Language and Linguistics UMIST PO Box 88, Manchester M60 1QD, United Kingdom. * Andy Way (away@compapp.dcu.ie ), School of Computer Applications, Dublin City University, Glasnevin, Dublin 9, Ireland. FURTHER INFORMATION For more details, please visit the workshop website: http://www.dlsi.ua.es/t4/ . You may also send a request for information to mlf@dlsi.ua.es . --=20 Mikel L. Forcada E-mail: mlf@dlsi.ua.es Departament de Llenguatges Phone: +34-96-590-9776=20 i Sistemes Inform=E0tics also +34-96-590-3772. UNIVERSITAT D'ALACANT Fax: +34-96-590-9326, -3464 E-03071 ALACANT, Spain. URL: http://www.dlsi.ua.es/~mlf From jeff.allen@free.fr Thu May 1 14:02:56 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Thu, 1 May 2003 15:02:56 +0200 Subject: [MT-List] Examples in commercial systems References: <2772.1051553512@lti.cs.cmu.edu> Message-ID: <04d301c30fe2$086702a0$4178933e@home> "Robert Frederking" wrote on Monday, April 28, 2003 8:11 PM > One other thing: in terms of examples, there are obvious examples in > research systems, although perhaps not in commercial systems. We might want to consider that Translation Memory (TM) tools are more or less the translation job production-based representatives of EBMT systems. Note: I want to avoid the word commercial-based here because some TM tools are in fact industrial in-house built production tools rather than commercial off-the-shelf (COTS) products. There are TM tools ranging from those that are more purely example-based (some in-house built TM systems), to those like TRADOS Workbench and other commercial tools that have fuzzy matching threshold levels that can be set by the user, to those that specifically allow for the Hybrid approach of using TM/EBMT + MT. Examples of such commercial TM tools taking the hybrid approach: * Atril Déjŕ Vu has a "Learn feature" which states "select the desired word in the source text box ..... DVI will ask the Déjŕ Vu Server to learn the translation for that word, and will display a list of target words compatible with the material stored in the memory database." That sounds a bit like a mix between statistical and example-based. * Wordfast is a MS Word macro TM tool which states to be fully compatible with MT commercial tools like Systran and Reverso (these are mentioned in the Wordfast docs). From what I can tell, this compatibility with MT is based on MS Word as the pivot user application (this meaning that Wordfast does not have a proprietary interface; it runs only within MS Word). Thus, Wordfast compatibility with these MT tools is when the latter are run within Word via their integrated toolbars icons (after installation of the MT software and accepting to integrate within Word). * The new T-Remote system focuses on providing remote access to information stored by any kind of TM tool (or by any kind of translation tool in general). Any kind of MT system, any proprietary system or any TM software can be integrated into the translation flow if APIs are available. In the case that APIs are not available, then querying happens by using the online MT portals. Here again, compatibility is possible by querying the various translation tools by a way (API, Web) other than through the proprietary software interface. For a good explanation of the T-Remove system, the next issue (#56) of Multilingual Computing & Technology has an article that will appear on the topic (Using Distributed Computing in the Translation Process by Philippe Mercier). Also see the software reviews mentioned in one of my recent postings to the MT-List. Jeff --------------------------------- Jeff ALLEN Paris, FRANCE Mobile Tel: (+33) 6 86 06 87 59 e-mail: jeff.allen@free.fr From gerbl@pacbell.net Thu May 1 22:49:26 2003 From: gerbl@pacbell.net (Laurie Gerber) Date: Thu, 01 May 2003 14:49:26 -0700 Subject: [MT-List] Article on MT deployments Message-ID: <3EB19666.7000606@pacbell.net> An article that actually discusses MT in practical terms, without either hype or fatalism! If only we could get them out of the habit of doing round-trip translations... http://www.informationweek.com/story/showArticle.jhtml?articleID=8900059 From jeff.allen@free.fr Fri May 2 12:03:50 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Fri, 2 May 2003 13:03:50 +0200 Subject: [MT-List] Examples in commercial systems References: <2772.1051553512@lti.cs.cmu.edu> <04d301c30fe2$086702a0$4178933e@home> Message-ID: <007b01c3109b$7e1cd560$589f933e@home> I dug up an article that indicates how TRADOS Workbench v1.07 TM system was integrated with MT systems (Logos and Transcend). HEYN, Matthias. 1996. Integrating Machine Translation into Translation Memory Systems. Proceedings of the EAMT Workshop, TKE'96. Vienna, Austria, 29-30 Aug 1996. To my knowledge, this system integration was discontinued a few years ago. Jeff ----- Message d'origine ----- De : "Jeff Allen" Ŕ : "MT List" Envoyé : Thursday, May 01, 2003 3:02 PM Objet : [MT-List] Examples in commercial systems > "Robert Frederking" wrote on Monday, > April 28, 2003 8:11 PM > > One other thing: in terms of examples, there are obvious examples in > > research systems, although perhaps not in commercial systems. > > We might want to consider that Translation Memory (TM) tools are more or > less the translation job production-based representatives of EBMT systems. > Note: I want to avoid the word commercial-based here because some TM tools > are in fact industrial in-house built production tools rather than > commercial off-the-shelf (COTS) products. There are TM tools ranging from > those that are more purely example-based (some in-house built TM systems), > to those like TRADOS Workbench and other commercial tools that have fuzzy > matching threshold levels that can be set by the user, to those that > specifically allow for the Hybrid approach of using TM/EBMT + MT. > > Examples of such commercial TM tools taking the hybrid approach: > * Atril Déjŕ Vu has a "Learn feature" which states "select the desired word > in the source text box ..... DVI will ask the Déjŕ Vu Server to learn the > translation for that word, and will display a list of target words > compatible with the material stored in the memory database." That sounds a > bit like a mix between statistical and example-based. > * Wordfast is a MS Word macro TM tool which states to be fully compatible > with MT commercial tools like Systran and Reverso (these are mentioned in > the Wordfast docs). From what I can tell, this compatibility with MT is > based on MS Word as the pivot user application (this meaning that Wordfast > does not have a proprietary interface; it runs only within MS Word). Thus, > Wordfast compatibility with these MT tools is when the latter are run within > Word via their integrated toolbars icons (after installation of the MT > software and accepting to integrate within Word). > * The new T-Remote system focuses on providing remote access to information > stored by any kind of TM tool (or by any kind of translation tool in > general). Any kind of MT system, any proprietary system or any TM software > can be integrated into the translation flow if APIs are available. In the > case that APIs are not available, then querying happens by using the online > MT portals. Here again, compatibility is possible by querying the various > translation tools by a way (API, Web) other than through the proprietary > software interface. > > For a good explanation of the T-Remove system, the next issue (#56) of > Multilingual Computing & Technology has an article that will appear on the > topic (Using Distributed Computing in the Translation Process by Philippe > Mercier). > > Also see the software reviews mentioned in one of my recent postings to the > MT-List. > > Jeff > > --------------------------------- > Jeff ALLEN > Paris, FRANCE > Mobile Tel: (+33) 6 86 06 87 59 > e-mail: jeff.allen@free.fr > > > > > > -- > For MT-List info, see http://www.eamt.org/mt-list.html > From pmercier@telelingua.com Fri May 2 12:13:19 2003 From: pmercier@telelingua.com (Philippe Mercier) Date: Fri, 2 May 2003 13:13:19 +0200 Subject: [MT-List] Examples in commercial systems In-Reply-To: <007b01c3109b$7e1cd560$589f933e@home> Message-ID: <000c01c3109b$d70e26b0$3c00000a@translate> Hi Jeff to my knowledge, this has always been a batch solution, nothing interactive. T-Remote Memory is the only solution providing integrated and interactive access to any TM or MT systems via the internet. thanks for refering to my coming article in Multilingual. how do you know about this? :-) regards philippe mercier telelingua software (t-remote memory) -----Original Message----- From: mt-list-admin@eamt.org [mailto:mt-list-admin@eamt.org]On Behalf Of Jeff Allen Sent: vendredi 2 mai 2003 13:04 To: MT List Subject: Re: [MT-List] Examples in commercial systems I dug up an article that indicates how TRADOS Workbench v1.07 TM system was integrated with MT systems (Logos and Transcend). HEYN, Matthias. 1996. Integrating Machine Translation into Translation Memory Systems. Proceedings of the EAMT Workshop, TKE'96. Vienna, Austria, 29-30 Aug 1996. To my knowledge, this system integration was discontinued a few years ago. Jeff ----- Message d'origine ----- De : "Jeff Allen" Ŕ : "MT List" Envoyé : Thursday, May 01, 2003 3:02 PM Objet : [MT-List] Examples in commercial systems > "Robert Frederking" wrote on Monday, > April 28, 2003 8:11 PM > > One other thing: in terms of examples, there are obvious examples in > > research systems, although perhaps not in commercial systems. > > We might want to consider that Translation Memory (TM) tools are more or > less the translation job production-based representatives of EBMT systems. > Note: I want to avoid the word commercial-based here because some TM tools > are in fact industrial in-house built production tools rather than > commercial off-the-shelf (COTS) products. There are TM tools ranging from > those that are more purely example-based (some in-house built TM systems), > to those like TRADOS Workbench and other commercial tools that have fuzzy > matching threshold levels that can be set by the user, to those that > specifically allow for the Hybrid approach of using TM/EBMT + MT. > > Examples of such commercial TM tools taking the hybrid approach: > * Atril Déjŕ Vu has a "Learn feature" which states "select the desired word > in the source text box ..... DVI will ask the Déjŕ Vu Server to learn the > translation for that word, and will display a list of target words > compatible with the material stored in the memory database." That sounds a > bit like a mix between statistical and example-based. > * Wordfast is a MS Word macro TM tool which states to be fully compatible > with MT commercial tools like Systran and Reverso (these are mentioned in > the Wordfast docs). From what I can tell, this compatibility with MT is > based on MS Word as the pivot user application (this meaning that Wordfast > does not have a proprietary interface; it runs only within MS Word). Thus, > Wordfast compatibility with these MT tools is when the latter are run within > Word via their integrated toolbars icons (after installation of the MT > software and accepting to integrate within Word). > * The new T-Remote system focuses on providing remote access to information > stored by any kind of TM tool (or by any kind of translation tool in > general). Any kind of MT system, any proprietary system or any TM software > can be integrated into the translation flow if APIs are available. In the > case that APIs are not available, then querying happens by using the online > MT portals. Here again, compatibility is possible by querying the various > translation tools by a way (API, Web) other than through the proprietary > software interface. > > For a good explanation of the T-Remove system, the next issue (#56) of > Multilingual Computing & Technology has an article that will appear on the > topic (Using Distributed Computing in the Translation Process by Philippe > Mercier). > > Also see the software reviews mentioned in one of my recent postings to the > MT-List. > > Jeff > > --------------------------------- > Jeff ALLEN > Paris, FRANCE > Mobile Tel: (+33) 6 86 06 87 59 > e-mail: jeff.allen@free.fr > > > > > > -- > For MT-List info, see http://www.eamt.org/mt-list.html > -- For MT-List info, see http://www.eamt.org/mt-list.html From cb@lim.nl Fri May 2 13:37:11 2003 From: cb@lim.nl (Colin Brace) Date: Fri, 02 May 2003 14:37:11 +0200 Subject: [MT-List] Examples in commercial systems In-Reply-To: <007b01c3109b$7e1cd560$589f933e@home> Message-ID: <20030502124030.FECQ1442.amsfep16-int.chello.nl@[62.108.30.75]> On 05/02/03 at 01:03 PM, "Jeff Allen" said: > HEYN, Matthias. 1996. Integrating Machine Translation into Translation > Memory Systems. Proceedings of the EAMT Workshop, TKE'96. Vienna, > Austria, 29-30 Aug 1996. In case anyone is interested, that article is online in the EAMT archive. See: http://www.eamt.org/archive.html All the papers for that workshop are in one single PDF file. -- Colin Brace | cb@lim.nl Amsterdam http://www.lim.nl From Christian.Boitet@imag.fr Sat May 3 07:57:54 2003 From: Christian.Boitet@imag.fr (Christian Boitet) Date: Sat, 3 May 2003 14:57:54 +0800 Subject: [MT-List] why no intgretion of MT and TWB In-Reply-To: <007b01c3109b$7e1cd560$589f933e@home> References: <2772.1051553512@lti.cs.cmu.edu> <04d301c30fe2$086702a0$4178933e@home> <007b01c3109b$7e1cd560$589f933e@home> Message-ID: Hi Jeff! 3/5/02 At 13:03 +0200 2/05/03, Jeff Allen wrote: >I dug up an article that indicates how TRADOS Workbench v1.07 TM system was >integrated with MT systems (Logos and Transcend). > >HEYN, Matthias. 1996. Integrating Machine Translation into Translation >Memory Systems. Proceedings of the EAMT Workshop, TKE'96. Vienna, Austria, >29-30 Aug 1996. > >To my knowledge, this system integration was discontinued a few years ago. > >Jeff As I underlined in my paper at NLP+IA-96 (Moncton), this integration cannot really succeed if the "lexical knowledge" of the TWB part (using or not TM) and of the MT part cannot be synchronized. What happens is that "suggestions" by the MT system stay terminologically wrong or incomplete after the translator has corrected or completed the TWB lexicon. That diminishes the usefulness of MT wrt TM. Of course, problems at Logos and its final disappearance are a "final" reason in the case you mention. However, Christophe Chenon, in charge of localization tools into =46rench at IBM, tells me that the integration of LMT in TM-2 has also never been more than prototyped, and one of the reasons is recurrent problems in managing a coherent multilingual etrminological database at each level (human and machine MT), and of course of syncronizing them. The only way out, it seems, it to manage a central multilingual lexical database containing enough information for human and machine usages, and to generate (or deliver on demand) appropriate information to human translators and MT systems, whatever their paradigm. The change needed in the MT systems is not a change in paradigm, but in architecture: the lexical modules should connect with the MLDB at regular intervals, and on certain events (such as encountering an unknown word or term or failing to build a complete structure=8A), in order to update the MT system lexical knowledge, in whatever form it may be. Such an "multiusage lexical data base" is being built by the Papillon cooperative open source project (http://www.papillon-dictionary.org/). The new version of Ariane-G5 on which we are working at GETA should have this feature of "incremental & server-based updating of lexical knowledge". Best regards, Christian -- ------------------------------------------------------------------------- Christian Boitet (Pr. Universite' Joseph Fourier) Tel: +33.4-7651-4355/4817 GETA, CLIPS, IMAG-campus, BP53 Fax: +33.4-7651-4405 385, rue de la Bibliothe`que Mel: Christian.Boitet@imag.fr 38041 Grenoble Cedex 9, France Mobile: +33-(0)6-6005-1969 http://www-clips.imag.fr/geta/christian.boitet http://clips/informations/plan_acces.html : carte de situation et carte de'taille'e / general situation map and detailed map ------------------------------------------------------------------------- Serveurs de dictionnaires: projet SILFIDE (http://silfide.imag.fr) et plus particuli=E8rement fran=E7ais-malais (http://www-clips.imag.fr/geta/services/fem/) Projet C-STAR (http://www.c-star.org/) et projet europe'en Nespole (http://nespole.itc.it) de traduction de parole Projet UNL de communication et recherche d'information multilingue sur le re'seau http://www.unl.ias.unu.edu ou http://www.undl.org, Projet PAPILLON de construction coop=E9rative d'une base lexicale multilingue et de construction de dictionnaires http://www.papillon-dictionary.org/ From lists@lissus.com Sat May 3 14:11:52 2003 From: lists@lissus.com (Alex Murzaku) Date: Sat, 3 May 2003 09:11:52 -0400 Subject: [MT-List] why no intgretion of MT and TWB In-Reply-To: Message-ID: <000001c31175$942bcc50$6501000a@Lissus> Dear Christian and Jeff, Indeed, integrating TM and MT has been one of the main goals of Logos = with the most recent(5-6 years ago) success achieved with TM/2. But = again, the dictionaries were separate so the problems remained. Big = steps in that direction were the various initiatives in which Logos = participated actively funded partially by EU such as OTELO, OLIF and the = most recent one TQPRO: http://www.hltcentral.org/projects/detail.php?acronym=3DTQPRO&topic=3Dres= ults We productized most of the features that were designed in these projects = and created some very powerful and nice tools for compilation of = dictionaries from different sources. Some of this workflow can still be = seen in the website of the latest Logos technology owner: http://www.globalwords.com/gwt/m2t.html In the same time, Logos had many ideas on how to integrate human = feedback from the post-editing process (both terminology and = syntactic-semantic rules). Unfortunately, events turned against Logos... Regards, Alex Murzaku -----Original Message----- From: mt-list-admin@eamt.org [mailto:mt-list-admin@eamt.org] On Behalf = Of Christian Boitet Sent: Saturday, May 03, 2003 2:58 AM To: Jeff Allen; MT List Subject: [MT-List] why no intgretion of MT and TWB Hi Jeff! 3/5/02 At 13:03 +0200 2/05/03, Jeff Allen wrote: >I dug up an article that indicates how TRADOS Workbench v1.07 TM system = was >integrated with MT systems (Logos and Transcend). > >HEYN, Matthias. 1996. Integrating Machine Translation into Translation >Memory Systems. Proceedings of the EAMT Workshop, TKE'96. Vienna, = Austria, >29-30 Aug 1996. > >To my knowledge, this system integration was discontinued a few years = ago. > >Jeff As I underlined in my paper at NLP+IA-96 (Moncton), this integration=20 cannot really succeed if the "lexical knowledge" of the TWB part=20 (using or not TM) and of the MT part cannot be synchronized. What=20 happens is that "suggestions" by the MT system stay terminologically=20 wrong or incomplete after the translator has corrected or completed=20 the TWB lexicon. That diminishes the usefulness of MT wrt TM. Of course, problems at Logos and its final disappearance are a=20 "final" reason in the case you mention. However, Christophe Chenon, in charge of localization tools into=20 French at IBM, tells me that the integration of LMT in TM-2 has also=20 never been more than prototyped, and one of the reasons is recurrent=20 problems in managing a coherent multilingual etrminological database=20 at each level (human and machine MT), and of course of syncronizing=20 them. The only way out, it seems, it to manage a central multilingual=20 lexical database containing enough information for human and machine=20 usages, and to generate (or deliver on demand) appropriate=20 information to human translators and MT systems, whatever their=20 paradigm. The change needed in the MT systems is not a change in paradigm, but=20 in architecture: the lexical modules should connect with the MLDB at=20 regular intervals, and on certain events (such as encountering an=20 unknown word or term or failing to build a complete structure=C5=A0), in = order to update the MT system lexical knowledge, in whatever form it=20 may be. Such an "multiusage lexical data base" is being built by the Papillon=20 cooperative open source project=20 (http://www.papillon-dictionary.org/). The new version of Ariane-G5=20 on which we are working at GETA should have this feature of=20 "incremental & server-based updating of lexical knowledge". Best regards, Christian --=20 -------------------------------------------------------------------------= Christian Boitet (Pr. Universite' Joseph Fourier) Tel: +33.4-7651-4355/4817 GETA, CLIPS, IMAG-campus, BP53 Fax: +33.4-7651-4405 385, rue de la Bibliothe`que Mel: Christian.Boitet@imag.fr = =20 38041 Grenoble Cedex 9, France Mobile: +33-(0)6-6005-1969 http://www-clips.imag.fr/geta/christian.boitet http://clips/informations/plan_acces.html : carte de situation et=20 carte de'taille'e / general situation map and detailed map -------------------------------------------------------------------------= Serveurs de dictionnaires: projet SILFIDE (http://silfide.imag.fr) et=20 plus particuli=C3=A8rement fran=C3=A7ais-malais=20 (http://www-clips.imag.fr/geta/services/fem/) Projet C-STAR (http://www.c-star.org/) et projet europe'en Nespole (http://nespole.itc.it) de traduction de parole Projet UNL de communication et recherche d'information multilingue sur = le re'seau http://www.unl.ias.unu.edu ou http://www.undl.org, Projet PAPILLON de construction coop=C3=A9rative d'une base lexicale=20 multilingue et de construction de dictionnaires http://www.papillon-dictionary.org/ --=20 For MT-List info, see http://www.eamt.org/mt-list.html From carl@iai.uni-sb.de Sat May 3 14:52:56 2003 From: carl@iai.uni-sb.de (Michael Carl) Date: Sat, 03 May 2003 15:52:56 +0200 Subject: [MT-List] Definition example-based MT References: <2772.1051553512@lti.cs.cmu.edu> Message-ID: <3EB3C9B8.3080907@iai.uni-sb.de> Eduard Hovy wrote: > > Bob Frederking gives a nice answer: > >-- an EBMT system is one that uses the parallel corpus at run-time, > > as opposed to a model trained in advance from the corpus > Yes, indeed, this is a nice answer. There is just a minor point which I was wondering how it could fit with the definition: it seems to me that also systems running under the label 'EBMT' examine and prepare the parallel corpus before run-time in order to have models for new translations. For instance, do you agree that one can think of aligning a bilingual corpus as some kind of generating models of phrasal translation units? If yes, is there any data-driven MT system (EB or not) which completely lacks this model-building pre-processing step? Many EBMT systems go even beyond simple aligned corpora and use more sophisticated (e.g. linguistic) devices to 'train' models in advance. This, it appears to me, is true for some of the older EBMT systems (e.g. Sato & Nagao, COLING'90) as well as the newer ones (e.g. Microsoft MSR-MT) including also the work of Ralf Brown's GEBMT. Ralf uses statistical methods to generate translation patterns before run-time. These patterns serve later at run-time as an analogy for new translations. What do you think are these patterns other than trained models for translation? Michael From ref@cs.cmu.edu Mon May 5 18:04:22 2003 From: ref@cs.cmu.edu (Robert Frederking) Date: Mon, 05 May 2003 13:04:22 -0400 Subject: [MT-List] Definition example-based MT In-Reply-To: Your message of Sat, 03 May 2003 15:52:56 +0200. <3EB3C9B8.3080907@iai.uni-sb.de> Message-ID: <5291.1052154262@lti.cs.cmu.edu> > For instance, do you > agree that one can think of aligning a bilingual corpus as some kind of > generating models of phrasal translation units? Good question. If you mean generating static subsentential alignments (word-to-word or phrase-to-phrase) in advance, I would tend to agree. I'm not sure I'd agree if you just mean cleaning up the corpus to ensure truly matching sentence pairs. > If yes, is there any > data-driven MT system (EB or not) which completely lacks this > model-building pre-processing step? Well, if you use the most basic version of Ralf Brown's system, it needs to have sentence pairs aligned in advance, and it does build inverted indices in advance, but it doesn't do any advance subsentential alignment (it does that dynamically at runtime). That might be as close as you can get in practice; I imagine that, to prove a point, one could build a system that takes a completely untreated raw parallel corpus as input, and cleans it up at runtime. > including also the work of Ralf Brown's GEBMT. Ralf uses statistical > methods to generate translation patterns before run-time. These > patterns serve later at run-time as an analogy for new translations. What > do you think are these patterns other than trained models for translation? While Ralf has built modules that use statistical methods to generate translation patterns in advance, not all the instances of EBMT systems built using his code do that. The ones that do are not "pure" EBMT systems, but as I said in my original note, I think nearly all systems that aim for practical use fall in-between the "pure" points in the space of MT systems. (On the other hand, until recently, SMT systems had no EBMT in them, since they made no use of the actual parallel corpus at runtime. Now that a number of them *do* refer to the corpus [to build phrase-to-phrase models based on the particular input sentence], these particular systems have shifted into the space between pure SMT and pure EBMT as well.) Bob From carl@iai.uni-sb.de Tue May 6 21:34:40 2003 From: carl@iai.uni-sb.de (Michael Carl) Date: Tue, 06 May 2003 22:34:40 +0200 Subject: [MT-List] Definition example-based MT References: <5291.1052154262@lti.cs.cmu.edu> Message-ID: <3EB81C60.5080609@iai.uni-sb.de> >I imagine that, to prove >a point, one could build a system that takes a completely untreated >raw parallel corpus as input, and cleans it up at runtime. > > sure! I don't have problems to imagine a pure (fundamental!) SMT system which trains all language and translation models each time it translates a new sentence. I guess you agree: even though this SMT system looks at the corpus each time, it does not become an EBMT system. Thus, we can think of runtime and compilation time SMT systems and we have runtime and compilation time EBMT systems From this I conclude that the distinction runtime/compilation time is probably an important detail of the system architecture but cannot distinguish major processing paradigms. But you also mention another point which seems more crucial : > build >phrase-to-phrase models based on the particular input sentence > > In EBMT, aligned reference sentences (segments, phrases ...) serve as models for new translations while this is not the case in (fundamental) SMT. Whether or not these sentence pairs are prepared at runtime or in a pre-processing step (and also how they are prepared (e.g. statistics, linguistics, logical form, LFG, etc.)) is secondary. Michael From ref@cs.cmu.edu Wed May 7 19:18:16 2003 From: ref@cs.cmu.edu (Robert Frederking) Date: Wed, 07 May 2003 14:18:16 -0400 Subject: Michael Carl: [MT-List] Definition example-based MT Message-ID: <18477.1052331496@lti.cs.cmu.edu> > >I imagine that, to prove > >a point, one could build a system that takes a completely untreated > >raw parallel corpus as input, and cleans it up at runtime. > > sure! I don't have problems to imagine a pure (fundamental!) SMT > system which trains all language and translation models each time it > translates a new sentence. I guess you agree: even though this SMT > system looks at the corpus each time, it does not become an EBMT > system. Thus, we can think of runtime and compilation time SMT systems > and we have runtime and compilation time EBMT systems From this I conclude > that the distinction runtime/compilation time is probably an important > detail of the system architecture but cannot distinguish major > processing paradigms. > > But you also mention another point which seems more crucial : > > > build > >phrase-to-phrase models based on the particular input sentence > > In EBMT, aligned reference sentences (segments, phrases ...) serve as > models for new translations while this is not the case in > (fundamental) SMT. Whether or not these sentence pairs are prepared at > runtime or in a pre-processing step (and also how they are prepared > (e.g. statistics, linguistics, logical form, LFG, etc.)) is secondary. So I guess the main point is that "runtime" is not really the right term; in EBMT the example-base is used *directly* as a set of models for translation, rather than as a training corpus for building some intermediate model that is then used to do translation. This sounds right to me. Bob From inge.blank@t-online.de Fri May 9 09:19:06 2003 From: inge.blank@t-online.de (Dr. Ingeborg BLANK) Date: Fri, 9 May 2003 10:19:06 +0200 Subject: [MT-List] question: MT system for Japanese patents Message-ID: <002801c31603$acdd40c0$0100a8c0@IngeborgBlank> This is a multi-part message in MIME format. ------=_NextPart_000_0024_01C31614.6C2DBE40 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hello,=20 I want to submit a question. The japanese patent office offers an = on-line machine translation service for patents (titles, abstracts and = descriptions). I am looking for information about this system. Which = kind of system is it? Which company does provide this system? Does = anyone know a website or some articles about this system? Thank you for = your answer. Ingeborg Blank Munich (Germany) e-mail: inge.blank@t-online.de ------=_NextPart_000_0024_01C31614.6C2DBE40 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

Hello,
I want to submit a question. = The japanese patent office offers an = on-line machine=20 translation service for patents (titles, abstracts and descriptions). I = am=20 looking for information about this system. Which kind of system is it? = Which=20 company does  provide this system? Does anyone know a website or = some=20 articles about this system? Thank you for your answer.
 
Ingeborg Blank
Munich (Germany)
e-mail: inge.blank@t-online.de<= /DIV>
 
 
------=_NextPart_000_0024_01C31614.6C2DBE40-- From inge.blank@t-online.de Fri May 9 09:17:56 2003 From: inge.blank@t-online.de (Dr. Ingeborg BLANK) Date: Fri, 9 May 2003 10:17:56 +0200 Subject: [MT-List] question: SYSTRAN use in the European Commission Message-ID: <002701c31603$ac927c20$0100a8c0@IngeborgBlank> This is a multi-part message in MIME format. ------=_NextPart_000_0013_01C31614.425F3A80 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hello, I want to submit a question. There is an article of Petrits et = al. presented at the MT Summit 2001 in Santiago de Compostela (Spain) = about the use of SYSTRAN machine translation system in the European = Commission. I am looking for more information about the system, = especially the lexical and terminological tuning. Does anyone know a = website or some article refering to this subject? Thank you for your = answer.=20 Ingeborg Blank Munich (Germany) e-mail: inge.blank@t-online.de ------=_NextPart_000_0013_01C31614.425F3A80 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
 
Hello, I = want to submit a=20 question. There is an article of = Petrits et al.=20 presented at the MT Summit 2001 in Santiago de Compostela = (Spain) about the=20 use of SYSTRAN machine translation system in the European = Commission. I am=20 looking for more information about the system, especially the lexical = and=20 terminological tuning. Does anyone know  a website or some article = refering=20 to this subject? Thank you for your answer.
 
Ingeborg Blank
Munich (Germany)
e-mail: inge.blank@t-online.de
 
 
------=_NextPart_000_0013_01C31614.425F3A80-- From Juan-Ramon.Del-Pozo@cec.eu.int Fri May 9 09:55:05 2003 From: Juan-Ramon.Del-Pozo@cec.eu.int (Juan-Ramon.Del-Pozo@cec.eu.int) Date: Fri, 9 May 2003 10:55:05 +0200 Subject: [MT-List] question: SYSTRAN use in the European Commission Message-ID: <91A2F218314CD4119566009027CA36EA0BBFB53C@ex2beimcombx04> I forward your message to my colleague Carlos Paz. He may probably help = you with this matter. Juan Ram=F3n del Pozo Translation Service European Union -----Original Message----- From: inge.blank@t-online.de To: mt-list@eamt.org Sent: 9/05/2003 10:17 Subject: [MT-List] question: SYSTRAN use in the European Commission =20 Hello, I want to submit a question. There is an article of Petrits et al. presented at the MT Summit 2001 in Santiago de Compostela (Spain) about the use of SYSTRAN machine translation system in the European Commission. I am looking for more information about the system, especially the lexical and terminological tuning. Does anyone know a website or some article refering to this subject? Thank you for your answer.=20 =20 Ingeborg Blank Munich (Germany) e-mail: inge.blank@t-online.de =20 =20 =20 From jaro.lajovic@mf.uni-lj.si Fri May 9 11:43:49 2003 From: jaro.lajovic@mf.uni-lj.si (jaro.lajovic@mf.uni-lj.si) Date: Fri, 09 May 2003 12:43:49 +0200 Subject: [MT-List] question: SYSTRAN use in the European Comm Message-ID: <3EBBA285.20110.FBA257@localhost> > Hello, I want to submit a question. There is an article of Petrits > et al. presented at the MT Summit 2001 in Santiago de Compostela > (Spain)about the use of SYSTRAN machine translation system inthe > European Commission. I am looking for more information about the > system, especially the lexical and terminological tuning. The EU-Systran has been presented on several occasions at the EAMT workshops as well as other MT events (although mostly from another perspective). Nevertheless, the EAMT web site - www.eamt.org - might prove a good starting point. Besides, there is an issue of Terminologie et Traduction on translation technologies/tools in the EU institutions with several articles related to the EU-Systran (i.a. Description sommaire du systeme de traduction automatique SYSTRAN, Resolving Ambiguities in SYSTRAN, and (partly) Probleme der automatischen Sprachverarbeitung). [Terminologie et Traduction 1.1998; Office des publications officielles des Communautes europeennes, 1998] If information required is not restricted to EU-Systran only, there is an interesting article 'Systran MT Dictionary Development' by Laurie Gerber and Jin Yang in the proceedings of the MT Summit VI [AMTA, 1997]. Regards, Jaro Lajovic From gerbl@pacbell.net Fri May 9 17:16:33 2003 From: gerbl@pacbell.net (Laurie Gerber) Date: Fri, 09 May 2003 09:16:33 -0700 Subject: [MT-List] question: MT system for Japanese patents References: <002801c31603$acdd40c0$0100a8c0@IngeborgBlank> Message-ID: <3EBBD461.8030009@pacbell.net> Hello,

I am not familiar with the Japanese Patent Office's system, but there
is an MT service provider in the US with an MT system
developed specifically for translation of patents from J->E.
The software isn't licensed for installation, there is just a
fee for translating each patent. You can see this at:

http://www.paterra.com/

It is actually used by some translation agencies for first-draft
translations.

Laurie Gerber

Dr. Ingeborg BLANK wrote:

Hello,
I want to submit a question. The japanese patent office offers an on-line machine translation service for patents (titles, abstracts and descriptions). I am looking for information about this system. Which kind of system is it? Which company does  provide this system? Does anyone know a website or some articles about this system? Thank you for your answer.
 
Ingeborg Blank
Munich (Germany)
 
 

From jeff.allen@free.fr Sat May 10 13:08:44 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Sat, 10 May 2003 14:08:44 +0200 Subject: [MT-List] re: question: SYSTRAN use in the European Comm Message-ID: <00b801c316ed$26462fa0$129d933e@home> Dr. Ingeborg BLANK wrote Friday, May 09, 2003 10:17 AM > > ... the use of SYSTRAN machine translation system inthe > > European Commission. I am looking for more information about the > > system, especially the lexical and terminological tuning. wrote May 09, 2003 12:43 PM > The EU-Systran has been presented on several occasions at the EAMT > workshops as well as other MT events .... EAMT web site - www.eamt.org - > might prove a good starting point. > > Besides, there is an issue of Terminologie et Traduction on > translation technologies/tools in the EU institutions with several > articles related to the EU-Systran (i.a. Description sommaire du > systeme de traduction automatique SYSTRAN, Resolving Ambiguities in > SYSTRAN, and (partly) Probleme der automatischen Sprachverarbeitung). > [Terminologie et Traduction 1.1998; Office des publications > officielles des Communautes europeennes, 1998] Evaluation reports, articles and conference talks on the EC Systran MT system have appeared at least since 1978. Early references are the most difficult to locate and obtain. Below is information for the most easily accessible documents. No guarantee that all the links still work. Mouton Publishers is probably now De Gruyter. Regards, Jeff ALLEN Paris, FRANCE e-mail: jeff.allen@free.fr ---------------------- >Envoyé : Thursday, May 08, 2003 4:27 PM >For more information on translation work at the Commission, please consult >the following Europa site: http://europa.eu.int/comm/dgs/translation/index_en.htm > >MT Management Team >Translation Service >European Commission >Luxembourg ---------------------- ALLEN, Jeffrey and Christopher HOGAN. 2000. Toward the development of a post-editing module for Machine Translation raw output. In proceedings of the Third International Controlled Language Applications Workshop (CLAW2000), held in Seattle, Washington, 29-30 April 2000. Order at: http://www.controlled-language.org ALLEN, Jeffrey (2002) Review of "Repairing Texts: Empirical Investigations of Machine Translation Post-Editing Processes", Multilingual Computing and Technology 13.2, 27-29; Available at: http://www.multilingual.com/allen46.htm KRINGS Hans (2001) Repairing Texts: Empirical Investigations of Machine Translation Post-Editing Processes, edited by Geoffrey Koby. Translated from German to English by Geoffrey Koby, Gregory Shreve, Katjz Mischerikow and Sarah Litzer, Translation Studies series, Ohio: Kent State University Press; abstract and ordering information at Information available at: http://bookmasters.com/ksu-press/ksu071.htm Official Journal. Call for Tenders: Postediting Services for Machine Translation Output. No. S 57. Brussels, Belgium: European Commission Translation Service, 21 March 1998. see: http://europa.eu.int/comm/translation/free-lance/en/ao-en.html Senez, Dorothy. 1998. Postediting service for Machine Translation users at the European Commission. In Proceeding of the 20th Conference of "Translating and the Computer", sponsored by ASLIB, October 1998, London. http://www.aslib.com (Reprinted as: SENEZ, Dorothy. 1999. Postediting service for Machine Translation users at the European Commission. In Institute of Translation & Interpreting (ITI) Bulletin, April 1999. pp. 22-24.) http://www.iti.org.uk Senez, Dorothy. 1998. The Machine Translation Help Desk and the Postediting service. In Terminologie & Traduction 1-1998, OPOCE, European Commission, pp. 289-295. More information available on-line at: http://europa.eu.int/comm/translation/bulletins/t&t/1998_en.htm WAGNER, Emma. 1985. Post-editing Systran -- A challenge for Commission translators. In Terminologie & Traduction 3-1985, OPOCE, European Commission. Habermann, F. W. A. (1986). Provision and Use of Raw Machine Translation. Terminologie et traduction. OPOCE, European Commission. Numéro spécial "World Systran Conference" , 129-43. For ordering the Terminologie & Traduction series: http://europa.eu.int/comm/translation/bulletins/t&t/subsc_en.htm Löffler-Laurian, Anne-Marie. 1996. La Traduction Automatique. Lille, France: Presses Universitaires du Septentrion. ISBN-2-85939-502-4 --- end ---- From tim.cavalier@rws.com Fri May 9 11:08:04 2003 From: tim.cavalier@rws.com (Tim Cavalier) Date: Fri, 9 May 2003 11:08:04 +0100 Subject: [MT-List] question: MT system for Japanese patents Message-ID: Dear Dr. Blank, For 9 years, I worked for Derwent information managing the development of their internal J<->E patent translation systems in Tokyo which were (are) based around the MT engines from Toshiba's Astransac/Honyaku systems. Our competitive analysis of the JPO's online MT system showed very similar (identical?) results for the same information although Derwent's terminology was far more developed. Also, once the JPO web site failed and it showed up an error which included "honyaku server" in the message. I can't confirm for sure - but perhaps you can draw the same conclusions as we did. Honyaku: http://www.xlsoft.com/en/products/utilities/astransac-e/transac.html Derwent: http://www.derwent.com/translations/jpmat.html I now work for RWS Group who specialise in patent translation: http://www.rws.com I hope this information is useful, Tim Cavalier RWS GROUP Translations Division - Head Office Europa House Marsham Way Gerrards Cross Bucks SL9 8DQ UK Email: tim.cavalier@rws.com Web: http://www.rws.com --- "Dr. Ingeborg BLANK" wrote: > > > Hello, > I want to submit a question. The japanese patent office offers an > on-line machine translation service for patents (titles, abstracts > and descriptions). I am looking for information about this system. > Which kind of system is it? Which company does provide this system? > Does anyone know a website or some articles about this system? Thank > you for your answer. > > Ingeborg Blank > Munich (Germany) > e-mail: inge.blank@t-online.de > From hovy@ISI.EDU Mon May 12 20:13:00 2003 From: hovy@ISI.EDU (Eduard Hovy) Date: Mon, 12 May 2003 12:13:00 -0700 Subject: [MT-List] Re: MT-List digest, Vol 1 #68 - 8 msgs In-Reply-To: <20030510121206.D18ED537E4@pairlist.net> References: <20030510121206.D18ED537E4@pairlist.net> Message-ID: At 8:12 AM -0400 5/10/03, mt-list-request@eamt.org wrote: >Subject: Re: Michael Carl: [MT-List] Definition example-based MT >Date: Wed, 07 May 2003 14:18:16 -0400 >From: Robert Frederking > > > > build phrase-to-phrase models based on the particular input sentence > > >> In EBMT, aligned reference sentences (segments, phrases ...) serve as >> models for new translations while this is not the case in >> (fundamental) SMT. Whether or not these sentence pairs are prepared at >> runtime or in a pre-processing step (and also how they are prepared >> (e.g. statistics, linguistics, logical form, LFG, etc.)) is secondary. > >So I guess the main point is that "runtime" is not really the right >term; in EBMT the example-base is used *directly* as a set of models >for translation, rather than as a training corpus for building some >intermediate model that is then used to do translation. This sounds >right to me. I don't understand Carl. What do the word correspondences in the tables in (fundamental) SMT serve as? (Say, the Translation Model in IBM Model 1?) And, taken as 'units', the distortions of Model 3 and higher? And Franz Och's translation 'blocks'? Surely as [models of] parts of translations, just like EBMT's phrases? The fact that they are composed differently at run-time into a translation means one has to differentiate the method of building the translation knowledge from the action of applying this knowledge during translation. (As Carl says, how the knowledge is built is secondary, though it is of course related to the run-time algorithm. What would you call a system that applied EBMT phrase correspondences, where the phrase tables were built up statistically? What would you call IBM's model 1 system if its word correspondence tables had been built laboriously by hand?) Perhaps you could define SMT as any system that, at run-time, composes its translation by combining elements (words, phrases, whatever) in numerous ways, and choosing the one(s) that provide optimal values for some (probabilistic) function that considers combination probability distributions derived from training data. In contrast, EBMT is any system that, at run-time, composes its translation elements using algorithms that optimize some juxtaposition/overlap function derived manually. And KBMT, transfer systems, and rule-based systems in general are ones that at run-time compose their translation elements by applying (typically, sequences of) manually built combination rules (that typically consider lexical, syntactic, and semantic features of the elements). the fact that these definitions are not mutually exclusive means one could for example follow the KBMT 'paradigm' using SMT. And surely that's correct. In fact, as SMT matures, it will be interesting to see if its increasingly fine-grained models correspond to the traditional MT pyramid of surface-syntax-semantics-syntax-surface stages, or whether it ends up looking like a more powerful version of EBMT's phrasal replacement and tweaking. Or whether the translation knowledge and steps become a new thing altogether. E -- ---------------------------------------------------------------------------- Eduard Hovy email: hovy@isi.edu USC Information Sciences Institute tel: 310-448-8731 4676 Admiralty Way fax: 310-823-6714 Marina del Rey, CA 90292-6695 project homepage: http://www.isi.edu/natural-language/nlp-at-isi.html From lorena.guerra2@mail.dcu.ie Tue May 13 15:57:07 2003 From: lorena.guerra2@mail.dcu.ie (lorena.guerra2@mail.dcu.ie) Date: Tue, 13 May 2003 15:57:07 +0100 Subject: [MT-List] translog and camtasia Message-ID: <3EA4FB980001E8CA@hawk.dcu.ie> Hello! I am looking for some information related to two program: Translog and Ca= mtasia. Does anyone have any information about them? Thank you.=20 From carl@iai.uni-sb.de Tue May 13 16:15:19 2003 From: carl@iai.uni-sb.de (Michael Carl) Date: Tue, 13 May 2003 17:15:19 +0200 Subject: [MT-List] Re: MT-List digest, Vol 1 #68 - 8 msgs References: <20030510121206.D18ED537E4@pairlist.net> Message-ID: <3EC10C07.10304@iai.uni-sb.de> Hi Eduard > > I don't understand Carl. What do the word correspondences in the > tables in (fundamental) SMT serve as? (Say, the Translation Model in > IBM Model 1?) And, taken as 'units', the distortions of Model 3 and > higher? And Franz Och's translation 'blocks'? Surely as [models of] > parts of translations, just like EBMT's phrases? OK, thanks for pointing this out. I try to make it very clear where I see differences between EBMT and SMT. As Bob says: in EBMT, phrase translations are build based on particular sentences. In this respect it differs from SMT, where frequencies of 1 (i.e. a particular instance) poses problems to estimate probabilities: an invariance which I see in systems running under the EBMT label is the sentence as ideal translation unit. EBMT has this in common with translation memories as has been frequently stated. But EBMT systems perform a (more) sophisticated matching and adaptation than TMs do in case all reference translations only partially match the new sentence (how this works: see for instance Somers' 1999 article). This sophisticated matching and adaptation in EBMT gives rise to the preparation of the bilingual corpus using all the available statistical and linguistic methods. Nevertheless, in EBMT there is (at least theoretically) a trace to an original example in the reference text according to which a new sentence is translated (that's basically the idea of translation by analogy) while in (fundamental) SMT this trace is kind of dissolved in the probability tables. Actually it's the opposite idea: in EBMT translations are generated based on particular instances (examples), while in SMT, particular translation examples are avoided (smoothed): in SMT, translations are generated based on probability tables which, in turn, are computed from the reference text as a whole. > What would you call a system that applied EBMT phrase > correspondences, where the phrase tables were built up statistically? I admit that borders are getting blurred in particular systems, but the basic idea in EBMT, as I see it, has to do with translation by analogy wrt. a particular (set of) example translations. > What would you call IBM's model 1 system if its word correspondence > tables had been built laboriously by hand?) Still an SMT system, because the idea remains the same. But it would probably be better to hear an SMT (wo)man. Michael From hovy@ISI.EDU Tue May 13 18:02:46 2003 From: hovy@ISI.EDU (Eduard Hovy) Date: Tue, 13 May 2003 10:02:46 -0700 Subject: [MT-List] Re: MT-List digest, Vol 1 #68 - 8 msgs In-Reply-To: <3EC10C07.10304@iai.uni-sb.de> References: <20030510121206.D18ED537E4@pairlist.net> <3EC10C07.10304@iai.uni-sb.de> Message-ID: Hello Carl, At 5:15 PM +0200 5/13/03, Michael Carl wrote: >>I don't understand Carl. What do the word correspondences in the >>tables in (fundamental) SMT serve as? (Say, the Translation Model >>in IBM Model 1?) And, taken as 'units', the distortions of Model 3 >>and higher? And Franz Och's translation 'blocks'? Surely as >>[models of] parts of translations, just like EBMT's phrases? > >OK, thanks for pointing this out. I try to make it very clear where I >see differences between EBMT and SMT. > >As Bob says: in EBMT, phrase translations are build based on >particular sentences. In this respect it differs from SMT, where >frequencies of 1 (i.e. a particular instance) poses problems to >estimate probabilities... in EBMT there is (at least theoretically) >a trace to an original example...while in (fundamental) SMT this >trace is kind of dissolved in the probability tables. Thanks for the explanation! You're right, I did not see it as salient that for EBMT the source of the translation knowledge is (or can be) a singleton example, while for SMT it is the 'average' of lots of examples. What is the operational difference? Obviously, one difference resides in the training / building stage, but for me personally this is of less importance. (You can build the same resource multiple ways, though not all of them will be easy or natural.) Another is in the actual translation stage: either having a translation equivalent that derives from one source (i.e., probability 1.0) or from many (with usually a smaller probability). The latter means its choice and composition algorithm has to be somewhat smarter. But again here, I can imagine a typical EBMT algorithm using the top-scored translation equivalents from an SMT-type table and composing them into a sentence. So I guess I continue to believe that the three translation styles (EBMT, SMT, and rule-based) are not incompatible, different, things, but can be blended into one another. But it is fun to think about the differences between them! E -- ---------------------------------------------------------------------------- Eduard Hovy email: hovy@isi.edu USC Information Sciences Institute tel: 310-448-8731 4676 Admiralty Way fax: 310-823-6714 Marina del Rey, CA 90292-6695 project homepage: http://www.isi.edu/natural-language/nlp-at-isi.html From Bruno.Cartoni@eti.unige.ch Thu May 15 10:45:52 2003 From: Bruno.Cartoni@eti.unige.ch (Bruno Cartoni) Date: Thu, 15 May 2003 11:45:52 +0200 Subject: [MT-List] unknown words in MT-systems Message-ID: <3EC361D0.D20F73EF@eti.unige.ch> Hello, I'm doing research on how to deal with unknown words in MT-systems (especially their morphological treatment). Can anybody direct me to some relevant references on this topic? Thanks in advance Bruno -- Bruno Cartoni Assistant Unité TIM Ecole de Traduction et d'Interprétation - Uni-GE 40, bvd du Pont-d'Arve CH-1211 Geneva 4 (Switzerland) Tel: +41/22/705 86 90 Fax: +41/22/705 86 89 http://www.issco.unige.ch/ From Angelo.Torquati@cec.eu.int Thu May 15 11:14:39 2003 From: Angelo.Torquati@cec.eu.int (Angelo.Torquati@cec.eu.int) Date: Thu, 15 May 2003 12:14:39 +0200 Subject: [MT-List] Reply to Dr. Ingeborg Blank about Systran use in the European Com mission Message-ID: This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01C31ACA.CB9C4D00 Content-Type: text/plain; charset="iso-8859-1" Dear MT-List, Herewith a reply to Dr. Ingeborg Blank about Systran use in the European Commission. Dear Mrs Blank, We thank you for your interest in the machine translation (MT) system of the European Commission. This is a fully automatic, multilingual system based on the SYSTRAN technology. Its dictionaries and programs have been developed and adapted for internal purposes. The system is used as a translation aid and administrative support tool by staff at the European Commission and the other EU institutions and bodies. It is also available to public administrations in the Member States. Please find enclosed an exhaustive article (in PDF format) with the title EC SYSTRAN: THE COMMISSION'S MACHINE TRANSLATION SYSTEM, adapted from a study by Ian Pigott, a pioneer of MT Development at the Commission as early as 1976. We hope that this informative article will address most of your questions about MT at the European Commission and the efforts we undertake to make machine translation as accurate as possible, both lexically and grammatically. If you have further questions, feel free to contact us. For the MT Management Team Sincerely, ---------------------------------------------------------------------------- ---- Subscribers to the MT-List who are interested in this article can contact me at angelo.torquati@cec.eu.int _____ Angelo TORQUATI EUROPEAN COMMISSION TRANSLATION SERVICE RL Directorate - Resources and language support Analysis of needs and multilingual tools JMO B4-70 L-2920 Luxembourg * +352 (0)4301 33960 Fax +352 (0)4301 34069 * angelo.torquati@cec.eu.int ------_=_NextPart_001_01C31ACA.CB9C4D00 Content-Type: text/html; charset="iso-8859-1"

Dear MT-List,

Herewith a reply to Dr. Ingeborg Blank about Systran use in the European Commission.

Dear Mrs Blank,

We thank you for your interest in the machine translation (MT) system of the European Commission. This is a fully automatic, multilingual system based on the SYSTRAN technology. Its dictionaries and programs have been developed and adapted for internal purposes. The system is used as a translation aid and administrative support tool by staff at the European Commission and the other EU institutions and bodies. It is also available to public administrations in the Member States.

Please find enclosed an exhaustive article (in PDF format) with the title EC SYSTRAN: THE COMMISSION'S MACHINE TRANSLATION SYSTEM, adapted from a study by Ian Pigott, a pioneer of MT Development at the Commission as early as 1976. We hope that this informative article will address most of your questions about MT at the European Commission and the efforts we undertake to make machine translation as accurate as possible, both lexically and grammatically. If you have further questions, feel free to contact us.

For the MT Management Team

Sincerely,

--------------------------------------------------------------------------------

Subscribers to the MT-List who are interested in this article can contact me at angelo.torquati@cec.eu.int

Angelo TORQUATI
 
EUROPEAN COMMISSION
TRANSLATION SERVICE
RL Directorate - Resources and language support
Analysis of needs and multilingual tools
 
JMO B4-70 L-2920 Luxembourg
( +352 (0)4301 33960
Fax +352 (0)4301 34069
+ angelo.torquati@cec.eu.int
------_=_NextPart_001_01C31ACA.CB9C4D00-- From rmk@iitk.ac.in Thu May 15 13:13:36 2003 From: rmk@iitk.ac.in (Prof. R.M.K. Sinha) Date: Thu, 15 May 2003 17:43:36 +0530 (IST) Subject: [MT-List] Re: MT-List digest, Vol 1 #69 - 8 : msgs: unknown words in MT-systems (Bruno Cartoni) In-Reply-To: <20030515103507.C6C81539A4@pairlist.net> Message-ID: Hello Bruno, One reference: o=09R.M.K. Sinha, Dealing with Unknown Lexicons in Machine Translation from English to Hindi, Proc. of IASTED International Conference on Artificial Intelligence and Soft Computing, May 21-24, 2001, Cancun, Mexico, pp 333-336.=20 RMK Sinha rmk@iitk.ac.in ----------------------------------------------------------------------- Dr. R.M.K. Sinha |=20 Professor =20 Computer Science & Engineering | and Electrical Engineering | Fax: +91-512-2590725 or Indian Institute of Technology,| +91-512-2590260 Kanpur 208016 India | E-mail: rmk.iitk.ac.in Home-page URL: http://www.cse.iitk.ac.in/users/rmk/ ----------------------------------------------------------------------- >=20 > Message: 6 > Date: Thu, 15 May 2003 11:45:52 +0200 > From: Bruno Cartoni > To: mt-list@eamt.org > Subject: [MT-List] unknown words in MT-systems >=20 > Hello, > I'm doing research on how to deal with unknown words in MT-systems > (especially their morphological treatment). > Can anybody direct me to some relevant references on this topic? > Thanks in advance > Bruno > -- > Bruno Cartoni > Assistant Unit=E9 TIM > Ecole de Traduction et d'Interpr=E9tation - Uni-GE > 40, bvd du Pont-d'Arve > CH-1211 Geneva 4 (Switzerland) > Tel: +41/22/705 86 90 > Fax: +41/22/705 86 89 > http://www.issco.unige.ch/ >=20 From donna@multilingual.com Thu May 15 22:48:24 2003 From: donna@multilingual.com (Donna Parrish) Date: Thu, 15 May 2003 14:48:24 -0700 Subject: [MT-List] Call for Papers: Localization World Message-ID: > This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. --MS_Mac_OE_3135854904_37682404_MIME_Part Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit Localization World is scheduled for October 14-16, 2003, in Seattle, Washington, and a call for papers has been issued. Organized by the Localization Institute, MultiLingual Computing and the Globalization and Localization Association (GALA), Localization World brings together industry novices, experts, customers and vendors to learn and share knowledge. It is comprised of preconference tutorial and management sessions, an extensive selection of workshops and discussions, and an exhibit hall of products and services. To submit a proposal, presenters are asked to describe in up to two pages the proposed topic of presentation, the points that will be made, the target audience and conclusions drawn. Submissions may be sent to submissions@localizationworld.com. The deadline for submission of topics is June 16, 2003. Suggested presentation topics and more information may be found at the conference Web site http://www.localizationworld.com. The Localization Institute, 4513 Vernon Boulevard, Madison, WI 53705 USA, Tel: 608-233-1790, Fax: 608-233-3511, E-mail: info@localizationinstitute.com, Web: http://www.localizationinstitute.com MultiLingual Computing, 319 N. First Avenue, Sandpoint, ID 83864 USA, Tel: 208-263-8178, Fax: 208-263-6310, E-mail: info@multilingual.com, http://www.multilingual.com Globalization and Localization Association (GALA), 240 Pleasant Street, Methuen, MA 01844 USA, Tel: 978-688-7200, Fax: 978-688-7222, E-mail: info@gala-global.org, Web: http://www.gala-global.org ******************************************************* Donna Parrish MultiLingual Computing, Inc. 319 North First Avenue, Sandpoint, ID 83864 USA 208-263-8178 Fax 208-263-6310 donna@multilingual.com http://www.multilingual.com ******************************************************* --MS_Mac_OE_3135854904_37682404_MIME_Part Content-type: text/html; charset="US-ASCII" Content-transfer-encoding: quoted-printable Call for Papers: Localization World Localization World is scheduled for October 14-16, 2003, in Seattle, Washin= gton, and a call for papers has been issued. Organized by the Localization I= nstitute, MultiLingual Computing and the Globalization and Localization Asso= ciation (GALA), Localization World brings together industry novices, experts= , customers and vendors to learn and share knowledge. It is comprised of pre= conference tutorial and management sessions, an extensive selection of works= hops and discussions, and an exhibit hall of products and services.

To submit a proposal, presenters are asked to describe in up to two pages t= he proposed topic of presentation, the points that will be made, the target = audience and conclusions drawn. Submissions may be sent to submissions@lo= calizationworld.com. The deadline for submission of topics is June 16, 2= 003. Suggested presentation topics and more information may be found at the = conference Web site http://www.localizationworld.com.

The Localization Institute, 4513 Vernon Boulevard, Madison, WI 53705 USA, T= el: 608-233-1790, Fax: 608-233-3511, E-mail: info@localizationinstitute.c= om, Web: http://www.localizationinstitute.com

MultiLingual Computing, 319 N. First Avenue, Sandpoint, ID 83864 USA, T= el: 208-263-8178, Fax: 208-263-6310, E-mail: info@multilingual.com, <= U>http://www.multilingual.com

Globalization and Localization Association (GALA), 240 Pleasant Street,= Methuen, MA 01844 USA, Tel: 978-688-7200, Fax: 978-688-7222, E-mail: inf= o@gala-global.org, Web: http://www.gala-global.org

*******************************************************
Donna Parrish
MultiLingual Computing, Inc.
319 North First Avenue, Sandpoint, ID  83864 USA
208-263-8178           &= nbsp;   Fax 208-263-6310
donna@multilingual.com     http://www.multilingual.com=
******************************************************* --MS_Mac_OE_3135854904_37682404_MIME_Part-- From Christian.Boitet@imag.fr Fri May 16 14:53:03 2003 From: Christian.Boitet@imag.fr (Christian Boitet) Date: Fri, 16 May 2003 15:53:03 +0200 Subject: [MT-List] unknown words in MT: article by Guilbaud&Boitet @ TALN-97 In-Reply-To: <3EC361D0.D20F73EF@eti.unige.ch> References: <3EC361D0.D20F73EF@eti.unige.ch> Message-ID: --============_-1159029179==_============ Content-Type: multipart/alternative; boundary="============_-1159029179==_ma============" --============_-1159029179==_ma============ Content-Type: text/plain; charset="iso-8859-1" ; format="flowed" Content-Transfer-Encoding: quoted-printable Hello Bruno, 16/5/03 At 11:45 +0200 15/05/03, Bruno Cartoni wrote: >Hello, >I'm doing research on how to deal with unknown words in MT-systems >(especially their morphological treatment). >Can anybody direct me to some relevant references on this topic? >Thanks in advance >Bruno >-- >Bruno Cartoni >Assistant Unit=E9 TIM >Ecole de Traduction et d'Interpr=E9tation - Uni-GE >40, bvd du Pont-d'Arve >CH-1211 Geneva 4 (Switzerland) >Tel: +41/22/705 86 90 >Fax: +41/22/705 86 89 >http://www.issco.unige.ch/ > >-- > For MT-List info, see http://www.eamt.org/mt-list.html here is another article: Comment rendre une morphologie robuste du fran=E7ais encore plus robuste en traitant finement les mots inconnus avec les donn=E9es disponibles Jean-Philippe GUILBAUD et Christian BOITET GETA, CLIPS, Institut IMAG (UJF & CNRS) BP 53, 38041 GRENOBLE cedex 9, France Jean-Philippe.Guilbaud@imag.fr, Christian.Boitet@imag.fr TALN-97 Best, CB -- ------------------------------------------------------------------------- Christian Boitet (Pr. Universite' Joseph Fourier) Tel: +33.4-7651-4355/4817 GETA, CLIPS, IMAG-campus, BP53 Fax: +33.4-7651-4405 385, rue de la Bibliothe`que Mel: Christian.Boitet@imag.fr 38041 Grenoble Cedex 9, France Mobile: +33-(0)6-6005-1969 http://www-clips.imag.fr/geta/christian.boitet http://clips/informations/plan_acces.html : carte de situation et carte de'taille'e / general situation map and detailed map ------------------------------------------------------------------------- Serveurs de dictionnaires: projet SILFIDE (http://silfide.imag.fr) et plus particuli=E8rement fran=E7ais-malais (http://www-clips.imag.fr/geta/services/fem/) Projet C-STAR (http://www.c-star.org/) et projet europe'en Nespole (http://nespole.itc.it) de traduction de parole Projet UNL de communication et recherche d'information multilingue sur le re'seau http://www.unl.ias.unu.edu ou http://www.undl.org, Projet PAPILLON de construction coop=E9rative d'une base lexicale multilingue et de construction de dictionnaires http://www.papillon-dictionary.org/ --============_-1159029179==_ma============ Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable [MT-List] unknown words in MT: article by Guilbaud&Boi
Hello Bruno,                                            16/5/03

At 11:45 +0200 15/05/03, Bruno Cartoni wrote:
Hello,
I'm doing research on how to deal with unknown words in MT-systems
(especially their morphological treatment).
Can anybody direct me to some relevant references on this topic?
Thanks in advance
Bruno
--
Bruno Cartoni
Assistant Unit=E9 TIM
Ecole de Traduction et d'Interpr=E9tation - Uni-GE
40, bvd du Pont-d'Arve
CH-1211 Geneva 4 (Switzerland)
Tel: +41/22/705 86 90
=46ax: +41/22/705 86 89
http://www.issco.unige.ch/

--
  For MT-List info, see http://www.eamt.org/mt-list.html

here is another article:
Comment rendre une morphologie robuste du fran=E7ais
encore plus robuste en traitant finement les mots inconnus
avec les donn=E9es disponibles
Jean-Philippe GUILBAUD et Christian BOITET
GETA, CLIPS, Institut IMAG
(UJF & CNRS)
BP 53, 38041 GRENOBLE cedex 9, France
Jean-Philippe.Guilbaud@imag.fr, Christian.Boitet@imag.fr
TALN-97

Best,
CB
--
-------------------------------------------------------------------------
Christian Boitet
(Pr. Universite' Joseph =46ourier)         Tel: +33.4-7651-4355/4817
GETA, CLIPS, IMAG-campus, BP53           Fax: +33.4-7651-4405
385, rue de la Bibliothe`que             Mel: Christian.Boitet@imag.fr    
38041 Grenoble Cedex 9, =46rance           Mobile:  +33-(0)6-6005-1969
http://www-clips.imag.fr/geta/christian.boitet
http://clips/informations/plan_acces.html : carte de situation et carte de'taille'e / general situation map and detailed map
-------------------------------------------------------------------------
Serveurs de dictionnaires: projet SILFIDE (http://silfide.imag.fr) et plus particuli=E8rement fran=E7ais-malais (http://www-clips.imag.fr/geta/services/fem/)
Projet C-STAR (http://www.c-star.org/) et projet europe'en
        Nespole (http://nespole.itc.it) de traduction de parole
Projet UNL de communication et recherche d'information multilingue sur le
        re'seau http://www.unl.ias.unu.edu ou http://www.undl.org,
Projet PAPILLON de construction coop=E9rative d'une base lexicale multilingue et de construction de dictionnaires
http://www.papillon-dictionary.org/
--============_-1159029179==_ma============-- --============_-1159029179==_============ Content-Id: Content-Type: application/pdf; name="TALN-97.JPG&CB.pdf" ; x-mac-type="50444620" ; x-mac-creator="4341524F" Content-Disposition: attachment; filename="TALN-97.JPG&CB.pdf" Content-Transfer-Encoding: base64 JVBERi0xLjINJeLjz9MNCjQyIDAgb2JqDTw8IA0vTGluZWFyaXplZCAxIA0vTyA0NCAN L0ggWyA5MDIgMjc3IF0gDS9MIDY5NTI2IA0vRSA4MjcxIA0vTiAxMiANL1QgNjg1Njgg DT4+IA1lbmRvYmoNICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAg ICAgICAgICAgICAgICAgICAgICB4cmVmDTQyIDI1IA0wMDAwMDAwMDE2IDAwMDAwIG4g DTAwMDAwMDA4NDcgMDAwMDAgbiANMDAwMDAwMTE3OSAwMDAwMCBuIA0wMDAwMDAxMzg2 IDAwMDAwIG4gDTAwMDAwMDE1MDAgMDAwMDAgbiANMDAwMDAwMTYwNiAwMDAwMCBuIA0w MDAwMDAxNjI3IDAwMDAwIG4gDTAwMDAwMDI0NzIgMDAwMDAgbiANMDAwMDAwMjU3OSAw MDAwMCBuIA0wMDAwMDAyNjAwIDAwMDAwIG4gDTAwMDAwMDMzNzkgMDAwMDAgbiANMDAw MDAwMzQwMCAwMDAwMCBuIA0wMDAwMDA0MTc0IDAwMDAwIG4gDTAwMDAwMDQxOTUgMDAw MDAgbiANMDAwMDAwNDk3OCAwMDAwMCBuIA0wMDAwMDA0OTk5IDAwMDAwIG4gDTAwMDAw MDU3NDQgMDAwMDAgbiANMDAwMDAwNTc2NSAwMDAwMCBuIA0wMDAwMDA2NTA2IDAwMDAw IG4gDTAwMDAwMDY1MjcgMDAwMDAgbiANMDAwMDAwNzI3MSAwMDAwMCBuIA0wMDAwMDA3 MjkyIDAwMDAwIG4gDTAwMDAwMDgwNDIgMDAwMDAgbiANMDAwMDAwMDkwMiAwMDAwMCBu IA0wMDAwMDAxMTU4IDAwMDAwIG4gDXRyYWlsZXINPDwNL1NpemUgNjcNL0luZm8gMzgg MCBSIA0vUm9vdCA0MyAwIFIgDS9QcmV2IDY4NTU4IA0vSURbPDY5Nzg2ODE3MTY3NmVl ZmZhNjkwM2E5MjBmYjk2NDM2Pjw2OTc4NjgxNzE2NzZlZWZmYTY5MDNhOTIwZmI5NjQz Nj5dDT4+DXN0YXJ0eHJlZg0wDSUlRU9GDSAgICAgDTQzIDAgb2JqDTw8IA0vVHlwZSAv Q2F0YWxvZyANL1BhZ2VzIDQwIDAgUiANPj4gDWVuZG9iag02NSAwIG9iag08PCAvUyAx NTAgL0ZpbHRlciAvRmxhdGVEZWNvZGUgL0xlbmd0aCA2NiAwIFIgPj4gDXN0cmVhbQ0K SIliYGBgBqI2BlYGBvYgBl4GBOAFyrAysDBwTIAKzIuVUPJ7du2Dg1Wwr7mv8QUpBQaG otM51wwYGALDWveudrp2RXd9V8SitVqBUa1bX0SH/ViYWxPgupYlisG93L2gvBxEgKiC ciC/HEy4F2AzFQiUGBjZLgFpUSCWBIuIMnDptV4MYTjHmsD0jUFfgOkXg64D0wUGXQOm m3yuhR7hYg0gZQABBgChxzejDWVuZHN0cmVhbQ1lbmRvYmoNNjYgMCBvYmoNMTcxIA1l bmRvYmoNNDQgMCBvYmoNPDwgDS9UeXBlIC9QYWdlIA0vUGFyZW50IDM5IDAgUiANL1Jl c291cmNlcyA0NSAwIFIgDS9Db250ZW50cyBbIDQ4IDAgUiA1MSAwIFIgNTMgMCBSIDU1 IDAgUiA1NyAwIFIgNTkgMCBSIDYxIDAgUiA2MyAwIFIgXSANL01lZGlhQm94IFsgMCAw IDU5NSA4NDIgXSANL0Nyb3BCb3ggWyAwIDAgNTk1IDg0MiBdIA0vUm90YXRlIDAgDT4+ IA1lbmRvYmoNNDUgMCBvYmoNPDwgDS9Qcm9jU2V0IFsgL1BERiAvVGV4dCBdIA0vRm9u dCA8PCAvRjIgNDYgMCBSIC9GMyA0OSAwIFIgPj4gDS9FeHRHU3RhdGUgPDwgL0dTMSA2 NCAwIFIgPj4gDT4+IA1lbmRvYmoNNDYgMCBvYmoNPDwgDS9UeXBlIC9Gb250IA0vU3Vi dHlwZSAvVHlwZTEgDS9FbmNvZGluZyAvTWFjUm9tYW5FbmNvZGluZyANL0Jhc2VGb250 IC9UaW1lcy1Cb2xkIA0+PiANZW5kb2JqDTQ3IDAgb2JqDTc2NyANZW5kb2JqDTQ4IDAg b2JqDTw8IC9GaWx0ZXIgL0ZsYXRlRGVjb2RlIC9MZW5ndGggNDcgMCBSID4+IA1zdHJl YW0NCkiJXJTPUtswEMafIO+wp9aZiYUs+e+thEImDAUGwqntwXEUUMeRXUui9A16Kbxb +0JdyTYhnWSi2PL+dvfbT56vJkdnDCJYbSdRDhQ/uOQZZEkGq92Ewv3kaHEbwb2eUEJp hE9W7l9Ecf/HJDhpdjuhDHRCbToBVgnYNV370NTNvRTQNWurjYCNhW1Xqt+l1NPVt0lI Sc45hxBxH3twNoLTHixU1SCwra1+pQgFpiulKTHhVirhM9dCY0qjQWKEUtYnSEnyFp6M 8Dj38PJRVD5wgxHPbpW6bZRc4z0XfnTGB03iXpMYojSGtEidKKEjcY/0tHNRqvD6Qday bQUs7pYX8+O7jyAMnDx0UhtZKphfLVenK8eOWI9kwFgCaR6PSMq8uMHidHU8g5OL5fXt DJYK4401sPx0vHDhjBROon1zPB2a40nhy/kS3J2fwTs4uby5/TL1cickTV+DfC6a7euf X0PCZ8BzGkewuDm9vJpfnEIlNuIJihmc4eAq4Tmc0NyNjb1yhpoPJCALK+t1aTcf5K68 J9tutheCzBtphBl3erEHA4bZQEZw4ppB8M2ztrvnw5nQXkD8oh2TgnmjulqKXomIxV6J z8FlY505lOkapWEaRoywoBosO/26OkcYKRA2tpMNNmQ05T0CDaf0tul2ogOLhIgUgRpW JMY8KFVZ/9TCdm+c/92KgZ8R9obP4pHP+2GZ7kVD1djHzpm6bZDims1JtI+K4mQ/LH8A RNePIyMpDpVE/42jn/DQwMFB+XNwUv7OoHovtAl/hRuJh60S9gm+2/eyxs4KNE2Axxl7 pRmJg1agC52EDC+GKvwmnmWnaVu/GLEXlvODxvnw4ogo649gW77Va7CXiznoh6ajXsXY Ti2eJFYv6lpg+Zi30dKMd1xBHCcjfKkxSYLNcycfy9cHCExx5FFwpfoyXTd9mf5FMaZL +2TWoKO1e60hLqeBRj+FQmmxW/tcPMXe11K47SRHXaqHRmqJV1mK6m2maR4I96OnX/8J MAAxuHpVDWVuZHN0cmVhbQ1lbmRvYmoNNDkgMCBvYmoNPDwgDS9UeXBlIC9Gb250IA0v U3VidHlwZSAvVHlwZTEgDS9FbmNvZGluZyAvTWFjUm9tYW5FbmNvZGluZyANL0Jhc2VG b250IC9UaW1lcy1Sb21hbiANPj4gDWVuZG9iag01MCAwIG9iag03MDEgDWVuZG9iag01 MSAwIG9iag08PCAvRmlsdGVyIC9GbGF0ZURlY29kZSAvTGVuZ3RoIDUwIDAgUiA+PiAN c3RyZWFtDQpIiWxUS3LcIBA9wdyBnZmUhxKgz2iZqiQLb7JRVqkssMx4SEloAsixDxHl bsmF0g1IHldlowbUdL9+r5vubncQgpXkwBkn3YddwYqiOJKuxxWXsPq5+0rV6WSetb8l OpDJkv2hbVhNl96ZAJtjyUo64zEXDb3oEE+5lHBsp/HeaXQqYPewLxvWUJ0M2X/r7na8 Yi0pMPcBk5cyJeeFrFNy9/tx0J5ctBt1CMoGAmEE5NfJkH6yPrjZYJ4YUbDja0RepnIw GH14sWo0P2Y9aoiz777valYn3yLnldERMw56HLVHp4OsmHzLEc8c4QJRnl8uUzgvAYMD 2Ml7cz8gZepJ9wTDqRCcuZ+DJ49OjaMKplfzM1HeT71R6MzIF6tJ0MMAX6esP00O/YBy haQXTNJlLwVraUhmiQV37xJ39YYp6+YWNRi/aOJnR0AhZdXw4jVsQJ+CM0HHyV3O0zA9 Iu54KkEl3tIZP/GgBLcToPmljI8HQO+mPfjDb70/1KyiNpmkgig3ZgFR1WZ66yLKSt93 Hz/dIibkd3PFMoTI3HLB26QasOeUgy86y1VdjHul7TQHMyQXwZqrljqW6ERfG2iVtHkr ae68oj1m+mK7itSugqIkJmhHBhQDyuP0NOhnUOcW9rKByvEP1t5e1wMMri3dljnw4sxT Elbj1AChIdsUWtAh79V2joHra07rDFjUskpx+2mEzjMhYoJZ3ZdwgY6zNf/BVVQi349D F3HdgCB/Up1xQrAVIPmcTApyhOdiw9DIVwHGKWT2+VWWNrFvLIypnf8m7uurcYpejVg5 qjNHqg/maUF+8K2wyQA7FQDplU9YqjeN065vhziuRN8s/RnmD+po4osRTbpbAoatDr6+ eW2TuRxu8rQQi1M4aBYfNRjBz3bVX8UJqGGFvQeThvtagM9s8TH6J8AA9btjCA1lbmRz dHJlYW0NZW5kb2JqDTUyIDAgb2JqDTY5NiANZW5kb2JqDTUzIDAgb2JqDTw8IC9GaWx0 ZXIgL0ZsYXRlRGVjb2RlIC9MZW5ndGggNTIgMCBSID4+IA1zdHJlYW0NCkiJfFRLbtsw ED2B7zC7SIDF8qNvdwHaAA2QnbsqsmAkOlEriS5JpekNuvIReqj0Qh2RUiwnRWFDtMWZ x3nvzXB3vWEZYUBh92GTUEJZJmBXbyjhLIXdj82XyKn6YWi/jwqaWDCSRyosYI7H2rRu NAqeB2162ak/GMN4STIMut1dbxKeQcIQf0anLIAzSvMJPbqR5pt+hIMeDcS7r5uMFKda OMvncE65D1f7vardqAzUysenpArxE/iCTqu08uHOyNapXg0O1PAWn2F1AZ+xLJA9SFNj LRIzujhhCI9Mwgqe0QuAZ7NoVdAy5PfawS+I85wUE5ZrkVYhCIuaOC0JRzC/4EtORNTJ AFquHaATlkelfK7qXo71g9pCrfteQaeNhTgRAotq4qpCvDGoLZDKv/R+LUZwawvSQ03W dsqC3O/bJ1xrPQyjBauHiUtjNKaBQkEk3EmroB18hMLNUBiZlH13xQFP22+wbyh+cCkY pDla2Ad30qmW6EY7m9Td0cJ7CHlizqMhD7+iIll2lpwG/y8H2f3EEnptDg+60/dTX25X fxUYfTdahy9XfBtkhc7YpXS7hcvdx6v/1S1ejs5WGn4anNHNWLtWD+fF85DMocwhnUan DwZUs5msyIpg5mfsJ44dMcwrWslTXGXghoNwxg6eZ0Y4W+gE3LVTK092c0pSlq/akRbz WWXhGyeyvw00rT1oq5bxOstgtDpxay7GMCOcYN+d2jFcCNE9DkUz7xer/WVGeZ6KQLBp vT6DbPFmwMuiitzcnpzQUgjs0BfwwoubLNX7EVSmV85h7qAHQEG6YGJ3MSuEgpVVtO/U 03SK6jo0O04qLCHq42nwJD7zqPW/rX+G/SySo7XtrJ54pR4XJy1WZyFhQbBqsaacLRdN uXA+mvZRuqUgpM0Y0maTvbd/BRgASFRuXA1lbmRzdHJlYW0NZW5kb2JqDTU0IDAgb2Jq DTcwNSANZW5kb2JqDTU1IDAgb2JqDTw8IC9GaWx0ZXIgL0ZsYXRlRGVjb2RlIC9MZW5n dGggNTQgMCBSID4+IA1zdHJlYW0NCkiJZFTBcpswEP0C/8PeAjNBg5DA4p5LM5324lum B9koiToYXCSlTb+glzr/1vxQdyVwaOoDQka7b/e9t9rdbrasVEJACbubTcnKsq5hd9jg 9vsm66/0oPtnZ/Ld100hJBP4g4LT2YIONyIeZnzLZYw4jMfT6Ky34zCYvjfXcBrDBBTf sqZptgmIgquSz8Gqodi7zE/aejNBlxcNqzOTFrjPi5bJTKflNy0iG9NuSAsQbv/iDYS8 qDAlfYgr5F92txvOWcvX4Lyslsqph4T+4uBh0kMHT+NB70Ov7WQgL5TITqM3g7emZ3GP eT/0c2bJtvUqr6re6LvXwSfmmjVzRFy19N7IuXkdnLPwLVxZzFw0kqnsFKxzJuFsWX2R KQK1K50W4gisekdzWceKMo2pfs4n1gWLeiGiTRLe28EcsVvojYM/x9G7GKXe+aRcpFdl nTqwwwFVD+6VwcfFOKR9UbNt1uVcsBY1jQvkyJvI9nppD/+NDL25Y8Xj2WssJ5VerWov VRtbI+UWniXJvHIoj2WiL8MTSuuvAel1V/rBeuiwv9heN2L6MBgYA5z64CyWHRMqxCpB gSgliAbBjlQRpuP0lVfxMy6ixDNNQ98jaOI0A+rPgaP0eu+QU9wRamcPcULIYI7F0rlC VQqOj8tsKUpyl30e4KSn3mCe8ES6dIZkmU6PYz8+WAPTuA/Om1fo9OCg19Bb78+T9mEy DD6NwcFADzvgv8ZhOfh+sOTkGudroDFTNFC0DFEOLKSqVzSux6Ve/PoD/ON5nGxsiFr7 z6jUhpRvQs6Vuusoplw5KllRXayoEshRWwe/yDNynmTxLqist0tQFTGIX/fs/MsRXyze Cmcy89mxiIkXgfz3Iqgu/lcz6E2kMec1Xjn4IEugAD7hN5dxj+5qU3AllUzBBH/Q1GEh WvR9QEL/CjAA62pqXg1lbmRzdHJlYW0NZW5kb2JqDTU2IDAgb2JqDTY2NyANZW5kb2Jq DTU3IDAgb2JqDTw8IC9GaWx0ZXIgL0ZsYXRlRGVjb2RlIC9MZW5ndGggNTYgMCBSID4+ IA1zdHJlYW0NCkiJhFRLbtswED2B7zC7yIBFiJREScsCBQpk1YV3cRaMRacMJNERRde5 QTeN79ZeqMOPHDmbeqGhqOHwvTdvvL1fpTkjDeUVpBS2X1dpRjKaN7Ddr3DV8AK2P1cP iVmnnNDkLYT4NrnAkksIfdiU4Q2kmaBVE6wft/cr3Ckhm+tnZSxfV7Urn4z6yZpJgtHu wPZltUimnPlsn3gU4176DEYKB9qnuZJFrEmLvAqQX+2d6hCCOWqDZ3hBmqS9swOuU9ok z2vGSJWMHl9N6qqpF5eyWYCC+YtF3ws1Suj1ePyhO/2sXm0AQjEr54uzGcK6Ap7Giwlp pFqQYjl3Ocle98fuMkmXkeYFyfG3aERW8siqotzX2yWTti/ajgYOnTwrPQyy6+QGganB oID2JIegIS0J57NEnlRxrdbUQaP2fVQnMc1lQK4pugEb6wM4eNqo6/fd+nOGVw9vmsm5 a7Lmg78TPPbrxgFRI9zkHzo9j2JogxR5wL7wZJGHEwVr8lBa7T0w35dd0oZ2VITekM5y Np8LlJ+EkQZpMFQbafgALVqCJXf+KfBJkdthEVmi/PMc92SMxu/u1j5s4ib6q+ZYNbjZ C0TR69jZ/5t6oWPGvaWjj8PnpYVoFKSkDQ/UplGchHJ9RD4VmluGEECU5Gaqyvl02XgU B/FbDyAPB7UXEQ5fGCj7wPvHiMGENpWf25SVnthNc8zfDWgL4iT30KL2N9/Ad358b62a DDw0j2hloQwchYHOJUt7hl/QCThoZUgYJbRSSpHOfGnh7vSrjAYpvosR5FnidOFsDNqa +EegcUiOOD6uIMrVWg8FhJ10j3OAQ71xQ+TuBvNmpksv4cuoxCDTb+XGdZYSnrTWK5pG JFdRy+vk/xNgABj2YhYNZW5kc3RyZWFtDWVuZG9iag01OCAwIG9iag02NjMgDWVuZG9i ag01OSAwIG9iag08PCAvRmlsdGVyIC9GbGF0ZURlY29kZSAvTGVuZ3RoIDU4IDAgUiA+ PiANc3RyZWFtDQpIiWxUy3KcMBD8gv2HuVmkvCok8TymKskh5725fFBA68jFApbATv7A F5NfzuiBYeNwYEAa9Yy6WzqQTvYP8kFBcno8CJqxooQUTl8O+GoOKeVZXcPp5XBHPp++ fruFp6Smgsz45kT7N4wJqzAqFwQxyf3p+6GgHJ+AdExpyiru8RwUuajJl2O0zrJsS0oF j0XLkoei7c3SGG0UtMpCwgUtiexl99uq2biBnFwGM/4cuuFBP80ux1VnBa3fN+LL51lE LrIyIE9G6kn2E3QJ45QRGQJseMpjHXlFcziydygWkNJaZAHp3Klfeuh71XXqFtrF6Gc5 rQOQHCtOK8dOQXMyheBGBa1JM1zGweppW+7Tyw/prhPOojLIVJltdHY38nnobVRQ4LPf eOUSyTzpTtvF5yB+vtOYjMNs4gTP95L52ch/mM+rPXiaRb1YybxexKm06ePWHAULWmwM piIyWOX1qsUfC80wPxsUxEI7g5mtddyx2juLZ54MH9x0UHnvr5RthJwR5k1q+9HTWLzw u7oT9zR6sLomrI76shT5fjc+jMp427rC+TWBbicI4XMl9q2hTZighRPRhbBI/EMeq/Ot Y29GZf7Pcl6uLMdD0Tmfo38yND/SpfsG7TMjbwlnSNDq25qy4pp4x7fvN00j1Fm+DT2c dY/WuzjKGjVNCtCUVv9Ay0wL9Ggv7A5GaWHB/2glBa9wHvoWHmf7NN/oRt9CIw14y5cV +TFHtgTfKeDa4PnKcBn1XyZXoEW5GW5KxQiRlbCdFWbvI77SX0Sp/C1RcWKH2R7jzRZG RrM02uLxlHhSkEC8LDr/Adcc2qF3nbDUN+IDuFx/uTSdRFs6q2Iuxlff2ulTaIcVoR2R Cfx6+SvAACauZ3ENZW5kc3RyZWFtDWVuZG9iag02MCAwIG9iag02NjYgDWVuZG9iag02 MSAwIG9iag08PCAvRmlsdGVyIC9GbGF0ZURlY29kZSAvTGVuZ3RoIDYwIDAgUiA+PiAN c3RyZWFtDQpIiWxUO3LcMAw9wd6B48bUzC6HH0mUelept7Nd0FquzYx+ISVnJoewz5U2 uVBAgvuxnUKCAIL4PDxoc0/7W/Nqxs6SBzpOQyBFwzSdi13NFPVRlHRCbUYtGy0aA2pb 1EixUxruD8b/WC0Ea0E5FEKCO7xi8IYezZN3cLwlhYY7PZwwSU3xuP+2kYIJISThZH+3 2XHGhdJk3204U2Vdkf3PzT19sn0gB0vSBcVq1TQXfy7L7K+qFv1n997Z8Puh2BJbyIoJ uqAg87R60tsAxUDIdUwRd7JkuiI7cQopZYoYg9HFG7fYwY4LKfbfNyWkUeoqO0dfJisu 0gUTgv1Fnv0UgrOehPV4dAtLl1sm6lp/bTXduzNjSF6SVc0pRcxQVbm/RmY8DrdmXTyg bQoh0vBANCh0hDdqDkWHRoPaglo+m1CMaAyobVFDsNtP7YrzbNoaa+mmYbAEcnJWxSxR EG+7F+vhOU8t2q9Qa9Wl8WFaQux7pyrWCsDnMgjOG8wnJNen3ifvnt1oyRuMZnx+95FW qgFCTSt81KwGBGSdek4iGVs6FELFxiV8h/ht4wsONUOyIhJR5JIVlnNpvrnihbe9s8c8 sE+cUC1WzRudUKIHmFVsM7l/CSuvwJj928F1ZnHHkPlWfnDm5Tm0xguR0SvA8d/QOnNH IHh/etstq7fEm9kd7F8Ciy40TGzNEta5KQEPCJhXo2HXm8G1POVvMxlvgAFQ9Ysdg5tG AgiXEesS1m1G0UeTS/HgV1NVVwVKcYEUuASbdoPgS8Y1gHdegjr3obRC33v5uIXya5hZ TU9/ho/rJYX6sMevxvX9CahPsCqVE/A6EzvEfYRZLNDVaBysG/BMCqCKjAgtKMAE20Ch 8eLxnwADAKjcS/INZW5kc3RyZWFtDWVuZG9iag02MiAwIG9iag02NzIgDWVuZG9iag02 MyAwIG9iag08PCAvRmlsdGVyIC9GbGF0ZURlY29kZSAvTGVuZ3RoIDYyIDAgUiA+PiAN c3RyZWFtDQpIiZRUwY6bMBD9Av5hbsVSQRiwgWPbrSpFqnoot909OImz8grsLLZ3tX/Q U3pp/7djMEm66qWJ4nEs5vnNvDf0m4QWOeW8gQL6myQr8oJ2DfS7BI9bVkH/ktym8nCQ O+cleJLRLm9SHSO4SSgnR6kdHEjGUxUWTe77TZKVTc7aqoKMrtAFq2fogJpaoS3sjNbi l5skDNLCVlhctQIBBzM5ZSaF536ycJzMcVInd7I56R8TmpcMcXN2Jk27SLpp6UL6Rh69 svCEtGc+7FJiUdRriVj5/LQ2Hm8Oi9LuNElrDdIjZZHX6Q9S8rxNgWDNbToQLIymYgkL Ng1ULvA8drBkLMJjndZNfoc1adgTnnepnFewr9b9HLHsv04X1DKUeYateUTtVtT+wzdM y6oSScoYl9Qqx85X/xC15G1MfvJiUO4EkqBOq2ZVTLzSrKiWTDyOmV6rZ4miGI/BvQc0 A895ejB6f7Jg/bTSv7q/biNK3dSz/F5LEFoMrxajd2YUTgWtdmbcKi2cUWiKWWu+2Oiq vWzBqsuyXhjdpWiQPcot0IjOeIf9HLCuMpXzaqMj27eOpNWZV0ANWIJks8EXkXE/LPvX sLfLXq77FAalH/Dimfw8B9EQV3wrvnaQ87n2o7FWbdHxdwQms/XWSWybelax4pgcctur idm/C00LT5Rv5a2rcxlxZlEKgZ188L+tmF2HvpaTQAc+I0bWlji+kqDFuqg9LSB8MTQU WmzLGAiwgl4IwH9+AtXPPSZ/wd8GwR8BVWvhJdzyFW7vC9gnFBQkeGVX5A2HMSkpC88s f4fke/KxT9qZWhuY4WFgFkildO7WhTeDpoq8z9NS1CwKC6Rm6PRPAmcN/TEaF94/I4px ClMfXkV+MQpy/iPAAJYUSDsNZW5kc3RyZWFtDWVuZG9iag02NCAwIG9iag08PCANL1R5 cGUgL0V4dEdTdGF0ZSANL1NBIGZhbHNlIA0vU00gMC4wMiANL1RSIC9JZGVudGl0eSAN Pj4gDWVuZG9iag0xIDAgb2JqDTw8IA0vVHlwZSAvUGFnZSANL1BhcmVudCAzOSAwIFIg DS9SZXNvdXJjZXMgMiAwIFIgDS9Db250ZW50cyAzIDAgUiANL01lZGlhQm94IFsgMCAw IDU5NSA4NDIgXSANL0Nyb3BCb3ggWyAwIDAgNTk1IDg0MiBdIA0vUm90YXRlIDAgDT4+ IA1lbmRvYmoNMiAwIG9iag08PCANL1Byb2NTZXQgWyAvUERGIC9UZXh0IF0gDS9Gb250 IDw8IC9GMiA0NiAwIFIgL0YzIDQ5IDAgUiAvRjQgMzQgMCBSID4+IA0vRXh0R1N0YXRl IDw8IC9HUzEgNjQgMCBSID4+IA0+PiANZW5kb2JqDTMgMCBvYmoNPDwgL0xlbmd0aCA2 NzIyIC9GaWx0ZXIgWyAvQVNDSUk4NURlY29kZSAvRmxhdGVEZWNvZGUgXSA+PiANc3Ry ZWFtDQo4O1c6LGdOKSgtJylfbjEiNV51TGRhLyxBSihaPT8tOyYsKVdRTmZXYy0/RjBf Kz00aDErKClJUzFqQWk4KUxhdQptSmghMCEwVC05V1x1OyUnLDFeL1xrZXFecHRPV1RT KGglOUVPV1w0NzBOWFc2QyJPK0ByXissN0gzOmtidG5BJgppa2BNJz8la004WG9jI0Fx ZiFocjlhP2tUT185UjZMSlxhMzMtIjcrNUAocjEpVm1QTTdiRS8uLldaaUdJXjNGaAov XmlTMEoqb0NHUTtKP242KmpGNj05YjRSZl9nM20xTWRgJVlnKHU3WyVwWkFFTUdUcWBu R2AkJDFvU1RNXHMrKApXT1I1OmlRUFxSWFpQP2RbVXBGL2J0R3MhU2shK1FNTyJcOCFM QFUiaVVHSllyQnBPJzEwL2NZPy5CST9eNkxDUQozODtCJFJobDZWSHRfR1M2W3E+UmRG ZG5IcHNEIWNAQGlAWVM5TDJISWpNRUZqVG5sPGBmQWUsZ0tmZSNENmtvKwpyJCNUP3Mv JS5QPVdBRlhcZjMpRy5hM1AkIzMkajBfbz4rcl5OPUsmZmw4YmpTT0EwSCpFTVdrYity TC1JYUYxRApBZzkqOjQ6PVFSL01FYE5NOVBPJW91cWdrNiQ4Z2ZZMkErN0deQ1wlLUEv LEZvSURoMidKcXVpQFsjcDtyVW5bVgo4UTZEMlJsR3NUcE08PVFCV0k6QkRBOVAzImhU UXM9TV50X2xtRGVfQzBgZiZJV2kpcz5UajwnYl4hSGFKXjBsWwpPK0JGVjZPWC9xPkZn XixTWm01dV8uXmMzVyVkMUwtTlBMck8pXzJybChCLFdNZ29hZ0lpNE4wI0NQVDYnIzgw RAolJlxPUEEvcTdSW0xUWz9BVitGKkstVHQjKFQzQE9fYz8sI1RkUldwLEM2OkIzLTNH LC8kdV40PD8kcz4lKDFqIwpZN1R1PS9iQCUwRm0+b1VtSWp1OEQ2IV04MlpoP21LYytF J0BaIVlzKSc0V1FdNG44JkBVM3VaYUhoQGBdaF81MQpecmkhNipxU01PUFtOKjtoZXRv ak9WbVgxOWBfSUFGLmRxaGYtamdAJ0AjTFdXNkJqQihicHFwJmhSKEYnbU0oXApdKlBf bz5DVydbTWhBcT02WHNHXkdFZCw7bSw6cG9cb0NsMDxzUl41ZFVFNnFiVkMyTFglOkRo Rz44cFFNdEVcYgo5ZkxZI0JoVzxgPzFOO3NbIit1SmBlPz1eRk5IMjZjbzhmclxvMy5A TEs8SFE4I2okKUo1PzxmP0lQaXNVUS5nVgpGR1tETEpOYChMYlZAXHI5T2ZLImZyLVoq aSJSKC4xME00ZmJqSTBxbUp1LEdiPjouckpAdCM6cCZgP0lcZzFaJwpGP1EpL1YnJlFA VGwtLFZxU2EqTFUwTz8pVllWJTExUU1nPGNCKVVrZyNOLmlOZF8vRVpmPGFpKCdCVUdI LCw7PQpfOCZOXW5EYzJzZHFrUmgmJUtKNjR0W09DU0xnSCMtPi9IIychUiU7TDJOLEZc TjNqUHEpRHJfRilEJHBUXSZgbgpaNSNJSms0KS4oTUlTQjJGcVkhUiEhQmFOR01YKyJj UkNfaTdOU1hbTyRHVEIrR0JxalJaLCZTLDQzMzs2QiU5aAosWltZPSQ+XCNWcUVyVD1M I1YnK287YDc0YWhxNF5SJmZ0NGBCTFZcXFY7I25oQlcnUnBEWipXUE0iRjk9TXFzMApN P0s1V2wvai5xTDtGRSQnI0I+SEVEN1dQbmw2cFAqJCw9NWYkQkc8IjZwPWRaWlxUMC0s PVZaXW9QZitWW2klTgptMzViRktFTUVHRDRycjY1bC9JTUBVKk5HS2swMFk4bixeWi9w Jz4tY1VZLVFDKmFiLCZqWTZkbitgalo8dERPQQpOMlxYQDJcS2wyKnBAPmowUCJGbXAs UyxEInJebnQmSF0kIi8qPSJLSFU1RVpLSTtwNlFMK08nTXBOODsnQzdFZApNby0+X0NL Vm9FWCNFRWA7OSVnZkBuVl1QXEhbT0xwJ2csOC9ePik4Oz47LChcKlFCayhzckROMFBw Tz1GVGg1NQpRK1M0VDRILDNYM2VUIVY8KUc+ajw9TGNaUF9BJDhtS1EmYGA7SDBYPTA8 aDFzNFFuOlw9PlheakZacDsoPTUmOApLQzZIP19bLClwKEcpMyk0O0phckVpZ1ZjOUxg K2RaJWkvcjFddUIhRjciV0ZBNVFxaVc9ZTAtaG1pa0xxNzxfSwo6U0E+JExQb0dkMWtV TVloITsvMlssJi5TTl4wLUdhT3VCbyUjJyxMX0VSaTFAPUtXUSo7cE4oKi1KN1ExOyYp UAo9VTQ6PlxNUFpvUVduUCckXChcImYsYUI8WnU9cjxaalgxJyRPRm5yOURDViJjQidX bFRgK3VpOzpHaXNuKVtfLwpWYycmZmpKPUpiWkAnMlBcSylIPU41TXVVLThjZUUtNnEi UTw/bD9fOWhjKDoyYyNMXS5gUWNGWEdjXDQ/TkQ4PApPJ2xjVENZJ2NfUyJWdF9gN19T UEtFVFpLUjJYMyo+Si0/Vl42cHF1NFQ3W0ZdKS5ETFNaV3I4LUxjXiJBTEM7OwppQ09Q SUk3LlZdKUdMWWoqU0dab2VSTSwyS0duNiZQZCZ0XzNgU08xOjM2Xmo4WyxMOkBJTExQ Vk8jIlFfNjgxWwpHQ1Q4UmhacyFzQlExQTE/UjVcNFI2aGZLKW9pMEkiajxuPGgyJTou Qj5Wbzc9LSM/UkwiRjIyIWNCSUg1QEBJWwo8dXI+cSE4L2E7MUh0Pjo3dFtCOypCaiJ0 aUJNZ2RfMyczKER0b20iZD01bCdsMW0lNS5RQUEyZkNccyZoRyFObQppIU45RkJlW3Mr Xm9SJFxxdFxqVTpzbmFBcUxEWGZUMmhwbkBbbjZuYVlXJlYoZnVwVGVUODMnZ0lMRzIz bkxWNApcbUUnOTlWY3BKTk9fXiUlWlpdbVgpXkohQiNBT2VuPVpdRWRGMj9qPlZaQjFD Py42PGVbSSxkTStycEA/bz44OwpzIVRPLGxPbHJcaGAxIlxmUE1ERywrT2FKTl1OJU5A UTNwLV43WzslL2FiS2hoQFlFbDVJLj5jQCxCWmchPzpMZApnKmkmT187bkxpLio6Mzc+ SGRbUC5sVkdoS1lVSEI8OSxecDtFXzZCXnFJImQ6SXJINkRJN01CMm4tJ0g8VW00YQpe bmlUQVRsN19yXDI9RkJwTDVXRDIjLm9qKyksXEwtND45NUYzPzRqKDhwXUBuUDw6LVkz cjB0bihDOnQ4Y0olLwpdWidNamRHYShmS09wMEJpMi5jXmoyQjhfbmY2OkJGT2tAOU1E J0osSz5SaEcyKGU9bDd1Wzo5XVReKTQ4OSxDTgorLUdcMkAlTCZPIy40YSMtZ0ZgP2pF cXUqSl9WIzNhXVtCW05oKSQ8ISlrPyoyJ29jPCI/RE4yajp1Rig0TCU4QwohRVFySVRV MGpbP0k7QHBgJW1rQSNwTygwW2NyMDQkW0VRZCg1ZV8yOmVsPU9PV2QiK2NxUTVhOE4h ZilmPTp0IgpzKWZRa0tMLkx0NCNIZF1PIVlwPUdrOzYlJSVEdVlETChsPzohNTNZR3Iy YW5wZ3FKamMvb1BPKzE7KUAqZ1YyOQpVRFAlcF1lbyt0b2Q4NGpIT0FZRVc6OE0uaC5W YkBJRyZOZmszOHRVPDY4RF5taD5BQ0lpO11OLV1aTkJbKGNpKgpZZFNNIUpTJltibUxZ NlpfMjomMi1LbithI0JQRDdqOi0ybTs0LkhMXDguXlVLQmgzPGx1PV0ubUo/KXFRI2M/ bwo2MW0oJ2NJVVc4SF9lTHVVUG5NZEFGb1lzPFBJYTk8MSwkKFFWKjBKLkhicGZcbkY8 ZVRFZkAmYlBQdTBUT3FeMApWL1FRUjQ5PUwxbjh0NlEwLDJUXTxYNGxYU3RWRzhrK1JR JkgjLjhHMGVMWlEwMztJYWNqc0A8UDpzLC9iYCdAQwo+bkspKENTMW5fTmRAWCJVJWgs Jm0nVjtAOEsrI3FoQjRDOi5YRGRQLz4rW1tRcj4yJDRrPWNlTUFzUXFRUDN1UAphRDY+ XDpSOzhPQ2ticCU5aEhsX1tRJTtZYi4iUTBaMztoN0xrVUdpaVlzaDhuKW8/RG5Ray1q NC9sbVRBOiIuKwpObUNQX2dtKTlgV1xWQlVtU1sxTlVAOGVEPEQ8UHFdLyJWdChfSm5p J1NcIWtldCUhNzI0a0glLS1qaVhSZFo6SwpkJVEyKW5TSUJdJEhvL0trO2YsYisjQWVl Z3InXF9iW0lVPVlsTEtxOU49Tis/WmFaTEEscmlPUDdkJEpcWTgzNApvUEVkRmJnNj1M XTM5QzRbYU9YWWJXP0JsS0dwM1omKz03N0w3O1dcKGhvJ2tYKGshcE4naStCPnVFR0Ay b29VYAo5Zk1aSW0waDI1P09oP1AtL0lkQCckNTRqczZrXS4jOUpac0pQSy51RHJvP0pp MFpWQ2FSPykxYWNhKktFQk0pMwplbm1SP1Vdb1JSJ1BSWWVVS2opRWU9RmIlLnBvPC9m NUg3cyo5WUJmWilfLihISlJpWVVMKV5QLnVUT1xFPVkrUQptOFU+M2g+bUNJb1ghXSs6 PURiNyw8UjFLXV1EclhESVg7O2xvXWdFZEJFTz4obVc1V2IhZiFzUFRaO3BhRGIkSwo2 I0ReP1NCLjElJEpYWkdQZFVaM0dXNDdXZzYkIi4mVlxKUShkRzUpKWRaWGw4ZjolKF5q WzlLYylLNFY8LENhXQooPDNtKEhvMDJ0UypSdEtsWVhNVWRRJz8uYSRcSUZTNERkWi4m YFlZalNzdWs2MCUxYzMtXWZNOmhtJT0nMiw/Nwo/UlR1YEkkK145V0RIQVdEPGNwcShf cGgjKGwvJyUubCksV0M8OmEnWSk/QE4tVzZuJlZ0LkNIMW4tJSxXKDxlbQpDTiE8WFg0 MCUpLUkwYVRsUCdYKWtVazFPbWNtXUZjX1tzJEw8OiZTXjJEPCpXOUUyPTx0QGosK0FQ czI9UyVKZwpIQi1AZm5lR1RlYDQwcElIPXBeJlpMNGRYK09PYz9qbV1YRWRacjQlS2ll X25LRVFfcl40PkRBT3F0akZKVWVgXgpAKUdycj83NzNBUmRsN1shJDpjczk5bi9OKlxQ X24xJVgral0pTU0oW0Bqc1EpczRiJHI/YW9YbDhQayInPSNqYApHN1NqOmFuZjNEJDpG IWBca08yUkA1a1dtJ2I0bDokYWV1XlROWTJYLTAhczlBL0BFNFFvPm46VDJiNl8/RiFb cwo+YiFnIWlsIVJOalYkPiElWXFZdWQzS1Q/XkVLNmkwRW1xPypmdDJRJWxOQjJPQyxB a0hHRz1yNiozZDxvb20hcgpAcy1eVV5VXHJhRlxRPDNnOyVzXE5jbSdYVUBwPiNyRStd TGNfRkNWWkM3MU1EJV5GMi5AIlA1a1MpSGBLNF5uNgphYi4oRUJIKDpqcG1mLihuZ2ks U1c2IzwmakE1VVtEVzZMcVYsPElsQmNCUClFS2AtLWUjMV9ZSUZgK1hWZSNgSgpPLjc+ SlRmXmFLSTYyZW9oO0QlR2NgWkxSZmQ8Q15icHMvJS1vbF9xI1tDVD1GQm4vIUNlcDQp NjwwdTJeMTUnLQpRazpjNW9naSpiKSMkZDRTdVo7Yz8vQztsZVVVbG9tPTY1UGQmMEZY JyMyPGZeSWk1WjNARih0TU9gdGowSFM4YwpMZTUwN1QiQihKUjhhUjo0QTghXCdKXCs2 ZD5qS1kjQVxhLyhVbGhOLVQ6TGInMz4iYSxSYz0lLGZlP0JCRFMjLwpWS0RRJEUhdF5K WlA8JCU9RGQ4RUdecHIqIic6VzE7b0BYaWRFTkBwbydbP0xaMl5CaShYMlwuRSEobFs3 JmkoXgo4PUdJbCtnaV44Q0Q6RjY8MWlKRE0wQFZNKzc9cGg4MlU+YlRcSXQ5SXROMSZe Vk4xNkgzTHBKaWFALHVhJUE4cApSVzYzZmd0MjspbT9XTi4zUVhaSWhyS0whSWUqbWlp dDVvbykoY10wMlhkZ1EsNiU6bDMlPW1jWSgoaDhwc1BiVgpjYWtLRThcJmJmWktyYEBd N0pXIkJyXUUmSGJNaTUwZDBPVm9PZ1tgWTBgXyQlUylnPlc6MmwjJjRLWGxlVmNyVwpx QWpGNVFKKm1AZWFZRXNhaDsmUUA9dVA6N1ZgQ25XV08qP0M1Pz4wMSRjLFIsLCZrNUJu Li9FVz5jTCcjVyVNXgpWNj9KXlMrN2drbWcqWm9rSl9KNVw1NFhJKGI7IkI9XTJiW11r YzpiMlRUPENWWk5lY2lTU00jS04hR2xcS002SAo3OCRBN1E3PFJvW202bWpGSHEnLEli Izo1Rz9aa3VUZlcxM3I4T0ZoUDxHSCJqUDE4UEhbWm4xTmlFKXQpKTRocwpSZ2R0PEtB SyJlb0c9bmMlcXFiamlzXj9OLyhtTGI8Xz0+dEBWZkJgWj8lLSUkIl1RNzY5NmUvbSdC OFFgMnM4KApHSjdkPSFPQChHRGFXLUhWMSNUJEVaakI8TmUxZi1dJkNoVyFDVGQ3PTNo VTQpNVUmSl46ZXFSXUhjbm0wPzo5RwpmYzUzQ04+KU9RS2gvUTUlXllecS9bPDthWSM0 U0BdRTo9I2ZORGQ8OFFdSGEmPmpaJXFUZnI7TkEoa1FxQVtqbAphSyxgSVFTTlQwLTFN LUIzR1RfOG0xMWpEJ0kkKzQ5bCY1JytGXnE7IzUtXENVSGM7KzFtKzxsRjVKIjMqQGEo NgoqMVhJdDI7OSpTQT05aEM7ZS5YbFcvY2k7YDYiJ2VNW2dYVUxyZ1Mock87NysvVSov VyhjbT1XRzE3cWNGNTY/Tgo7UE5NUVVGXUBqVlkjUTBIUWtnKnJZMzdhVTs5aHAxPTtv cUtvSGlmUEs5LlxLcSpNK25yNUxUTEJVaUsqcmZEJApWdDgpRThjU2BvJyModXFpUzpj QCU6bVc/MT0yWVkhcVMhV0tUc1tELiE8Z19wOXF1NTBZYVdTU14sWzZnUCFtKQphb29h JmdLNmo4PWc4TmlIQkZiJFtaLzFnO24oQ0VbKzVhLGp0LSNTZTRTLHRoP1BHZzQsW1tZ LDZyLzdrPWxnKApacHErbjhyS0xRO0h0XSFVbzBNcC9SPzJZXCNQJXFqOWZEXUFCMFUh UltyPl5ab1FcZys/VWxIKEIpLE4kUDNTUApyI2RrWFY8MEklUmlEaDNXa0IoL0ZgWVpQ TUhsU081MWJpdDBeQ1RDbmF0N2MsPyxHXSZNSCdUSDVFOS0jRmRYIgpLbyQ8UmI2WSlN cTVMRFw/R2U6L3JsSyxCNDVCXVA2YjJyYSkoYSJsX0tFRFQ+J0xAJnBPciwsXGFqTEE0 SCFHWwpSJiozJTInOChXJHBmXE8oMjMpVUBlQUosP2I0c2NfVnA4dGlqQEEuVVt1PS5B Xjo3V1NRdDcwTjVdQ1BPaWZEVgpeM2hCcTdXSl5XNWlvQ1opXVZCOzRKRVxlXm1SVipY L1pZPyRSdCErISo9YVVPJCJha1w2U2NTaDxGVS0/Skc/LQpdY1FmbGQtRmw9NDIjUClI R0BpTFtlXipcInBnKjlnWCcrWmRUJmdcSW1rLGcrbj06Rl09IlpaTTJgSUI2NEY+XApl LjMhNzM+NkVeYmhQST9BIikmZ2BFIUpjLy5xMklTVlc5KmVxMlgtOW1bXmckUlZoSyhN VC1QXWNybDFqWjMlWgpqQDM7NCJZKlxdTVk6dHFJbUFfMWlyaEQ4RFstbTkrM0JJQExE MFBUSExqZS9VZFtETixHJUhFTDtFTyFJUkJIQApAbCE2WzddVCdqYlM3RmNYTmhrJTY2 JVtzWSZxVV9GLj0iZz0/RDx0ZXRqP0U4T1YrVzUpazdeMEBDP1QnLk5oWQpJXWJpJCpj KVdUJU5RRmVhIVcpMCppL2hAQ08rQ2NyUVRocERyODQhJSNPPHA+VlgoVSU7WzBHLClA bSNoMzhkLwo7UGttVEorZj9ibV9kbT8vJjVtJCJtUnR1UnI4SmBCW3BXP1pTWVpNOGY2 cFVDIkZOTkFAWlItZy49S1ViVCVfawpZbWYyKy0iYl0hRipBLD9qLlNxcCtOISFGTzJD aGZlRTVpbDpxLlVEJltmcXInYmUoI2U9bmYuY3MkZiInYEVMRwpzNDo1R1JmcTdVXlwm PSU9VE5GIlFbUjMuLUs4aF1FPU9SNic3RlpQVT5vJ1lwKGtoPjVuQm1aVCQ9MD5nKnVR UQo6JjAlMnJiMXBcTVtcbDBRZDlccSdQWTQpPFY/cyldN1pyNyhnKGImNUldLkokXF1L L1Q9WG0qZSFhX1kxSDEoJAosP1o3LWpOLjhXUWVPSU5HSlRPXDhCbyIvMTtZSWZSUmda S2g3TTs1b1lnRjdUNFJkRD9tLmRhbzdRZjNeVjEuYgpBMTY2LChkSz1rYmZIMWlLZzJo IjxUTUFyMCxwWDxUImdXaV4jZV0xcmZsbzpPSUsiYShvOyUsXDxsNXUrTE5SNwotT0Nt LnJ0NCkqI0NkPzRwNjxVQmUrUFpFVktdUWFDbVY2U1BUW0VLL3JAUVpMU2ZNTzpFPjtC Y2FQay1mXzgyPgpiJjtLPy4kZEMiJ1A5QCtNciMrPTpIbVtpSSlJYDxlZEcsKDtVMTdw NkdgUCttZkspTCVkIV1JTVU5a0BNYVcjJwo3QS9YcGw+b1okOV47LSttWVhwIU02PG4/ KzNYXTQnXEFdclw6OGAkayddTSI6VkwuQmM0ZkJkND80a3BLISgvTgpXN1dSX0ZGRD4h b0RBb09ZaT5SXyFvR2xGUGJ0cz1GKml0V0RZXi00bixOKFwhPEFCQzxRPn4+DWVuZHN0 cmVhbQ1lbmRvYmoNNCAwIG9iag08PCANL1R5cGUgL1BhZ2UgDS9QYXJlbnQgMzkgMCBS IA0vUmVzb3VyY2VzIDUgMCBSIA0vQ29udGVudHMgNiAwIFIgDS9NZWRpYUJveCBbIDAg MCA1OTUgODQyIF0gDS9Dcm9wQm94IFsgMCAwIDU5NSA4NDIgXSANL1JvdGF0ZSAwIA0+ PiANZW5kb2JqDTUgMCBvYmoNPDwgDS9Qcm9jU2V0IFsgL1BERiAvVGV4dCBdIA0vRm9u dCA8PCAvRjEgMzUgMCBSIC9GMyA0OSAwIFIgL0Y0IDM0IDAgUiAvRjUgMzYgMCBSID4+ IA0vRXh0R1N0YXRlIDw8IC9HUzEgNjQgMCBSID4+IA0+PiANZW5kb2JqDTYgMCBvYmoN PDwgL0xlbmd0aCA1ODg2IC9GaWx0ZXIgWyAvQVNDSUk4NURlY29kZSAvRmxhdGVEZWNv ZGUgXSA+PiANc3RyZWFtDQo4O1gtRGdNWWI6JylfbjM2Uis3WSJyPitqT3BuUSpbYzdL KGZUJSsqXyNpWF81YDlhLkpPRjZ0YWEhaG5ULzluLwpobTwzRz9lWGNhZnFiRmJmVEA7 PWJwSm9bVnM5ZVA1JHNobUlSYXU+TW9WInUibjdLYE0paDpFL18yOU1xS2hgLwpyPmVA N0NILFpQamtoLi0qaD1wNDQyRj5AJVltVEo6XzdwUC9GXDk3LUwrPmMkWSJEWTBBXWlp Km5xJDUlSmkiZQpCNkBFaS11RGdbLSRiXVZaX0JIS1glQ190alo3KjJabi4tSFdTNSFO ZG9kQ11lOCh0Z2ZJKCIhJXJhREdeQDhqOwphbFkmJCxPNCozXlVIZDxqaWFSJT5jWG9W UVctdWFCOWg6S2VJWlVqYkcoRihdKGc1Il9gbSRtIyM7b2oqaileZAouQXFRSjcwbWVJ TixoQW5tMGs8Y2RhIUBgPUtfQVIqKltKWGVXRy1iM01JRXVMU2NtMlVYZXA3M2c0SEpB WCwpNAozMnMmS0AvO0BWKWZuQFBiQ2oxKzU7OE9fbkgxU2VJSTpyRSEpbFReSD1dKS9l bihXJkktYEgub3FuaCJbO205LwokLCouO2JAU0U0YlNuXHVpIltqQVNoMiktRFtoUERT Q2Y1WlM4LSsxRW4pJ2krcHI3JldqO1ItOTpoa284aVczbAprUTw9NzZJTXMyPi9JQ15H dGBZQ0B1PFdrMEZcKWRRZDFMR3MxTytbNWNaPmdhTjZYbF1gPyhwbmw9T19ic0RDZQpJ JTlzY0RVIixyPzdWO0Uvcz1jImxISG80TnVsLm1bS1s7RV9PbDUhUT11XGdDYDUiIjgr bTY6bidgbV5LUSs3IwokUyQoWDNTNzwhXU9PRywlXGYpc1ZkKGsvV2xnMkZYTDUlOWJB QFQoQj0zZicxMVlGOkh0XWdbNzRRTVVmYEVSaQpIN148PS5zNipMLUA5WFNVbFU5aCJX IU1uUkstckJLPGgzVUMxV1FVN0skPWhsUmNALjNldDhmRD0wXlYqbCZBUQopTUlAPypE Xj8iUVdDIWJlbmAlI0A2TzZeWTA1J2dQbFVIW2Nva281ZWA3VUpTInJUPS5RcDReNUVQ NTk4YithYwonS2lSLE9BOEteZydYLXJKKUdyUSNHJlcnQTNubzY6XWReQ2tUZU5ya0VW IiRLalRkdGBaImw+TTRJJ0JtRj5hPgo4UlVtUD82aTc1PTQvRiZWZ3FZVmFRa0dsRUA2 Kyxgb0pVODxPMGdWXFFpP1A5W09vU0pEZFskVkNZV2xhIkg7VgowRiRPOiExdWs0akYu TlM1VV9gJ2QwRnFzXmk5SDFCKlAhRCZNVUdDN15SczZlKCRaMFQ8OHQwPllLTVMnJkVZ RwpgS2hBcmRuXFthU2opOWU4OjYiVERnR1Y2PkxiZkYjdDwwMkMnYz5mayNWQTxOMm5G KjNEOW43OD9Fa2JaKF5ZTQpXKllaZmxiJj06WzdNaHMkVz09aDpRZjglOjMlWklSQ3B0 SGI2ZUFmJmU1cikuaT8+RUwzTjZxX187NCEvN1huJApHQF8vPW48PklJT1FlJCtUMi88 ZSs9MlQsO0VKIk9qMkshVSQiYTdpOEVYOXU2V0NWRUhtQDVkS29vJ3M9aWRDago2NHQm MSpjJ2xWaVs3WiRIZVRZNlk/UVdobXAvMGVrKTtcWG9jUWpxT1pmYVU3RVRaSDVXR1ZO KCZjIy0pREZvPgo0QTZuPFIxXWRnQkcwKHAkNG9tTiJMWWJkI0U/RDZdYHNfUz4vc1NL b2MySzByOitiN2ZEbGM9W1ZmL0BkTVJoOwpfWypnckZRQ3MxQGxnPT0vOCJhPFBmaVBL L1JGZG9rLERVTj9fc291Rj1LQS85YDUoVnFaJSkjUWhmal5eYGpkRApiZik0JG1CTXJy b1FZO11aOiFwRldSOXAsbEppP20/K2lLWFUkbF1sQFdaMGczRCFbY09hOEMlSTkzZ0Fl JkVlOQolXW90LjFcPDglLjAwT0FMaCg6MGcqUTA4OTwoKyJkKE1kdFMhPklVPiNbai9q QWs1WDBedGtnUjkmLjpOPDtoPQovM3IxaGhQVSVyO2shZyo7QTw0bFRwSy9xUVZgLHRP NDtgOi9XQXAuOTdiNFchRUUmQilaX05FPlg4JjZmMDZPNAo6MGMuSVNIbSJaUF1nbzdY TkVGMVlgbFRLKFVrKClyQCctQHA9QWUiXU1dZ2lJMV0zNDhfSE8oUSdJUGRXYSIiXgpA a1ZUN1FrV1hNVCFoaTM2X1dQTiJkTVxROmJvIVM+NWVaYmZRaTNuY144U2YzYW06aTZx YGhVLkljNkY7N0UhNwpTPTRHN2BYa1NoQXVhWDBGMyF1XUhtbUNab1FHbUQiM2E8RFZC TiVwV2RpSj87VV1vNmtXMFddXS9bIlNaSVdpagpyJzlPQDhsNyssMCQ+OmklInVxc1Fk PT8yIixaXlVyPElibE9vZFw8JScmMjc1bkBON2YzdFFXa0dxIUtvIlJlbgo/YmFMKVA3 amojN3RhM2ZHVDxZVk1CaTsxa3QqYXBuL0duUSo8X2sxSlEuIjZjbzdMcmhnNWI5NUNM JShdRDZKcQoyaiJFamgsc2MyZ1xLc1VnXEtydGNZPkxnLEU5Nj01aztVRGhaa2pzMlp1 QiRvRVc1Nlxqakk0KVBIXipVRSNoXAooajY/ckE7SmdtM1Y7OzRYXkRGdGQoVyIoJ3Jv PSFFIT0uLk1GMFZxY0gxKDhBMmlgJ2UzaFd1JV5vL0pXazI3agpjZTZGUFBPbGBuZVZJ bTw2IitrWmddSi9pK0JPbjpjSzVMZzdkUF9cIXVSU1lxdWxVTW4najI2SzZcPmdiSmZn Kwolcit1SVEwWSpLZlZhakNLcCNObExKVDRTXC4+J2NKRnRZXnJLU0NlbFlkJCY0MVhC aWFEO1c3PCRZXk0uXjVAVQpRWiklL0RrUSNzNSFXTTtSODx1UWM+RFAsJ2wpcklpIVxP ak1RTmRZW2ktOyciNFRbNV9SI14+S2Rtb05nJHI9RgpqTTFtMkphRiQxXjFyQ0cnK2Mp RDdDXERJVTooLzhpJzwsaDpeL1A6cF0tUGFnb2dBViY2TD1vYCE7ZG5HOFM4WwpvUFlM ZXBndUJyI0JrLk5icWBbQj8pZDIxUSducTZJZV0+VW9ObDM2amA1UXVQU3RbYEg/PTph PCdtQSdLZmlHQApPImMyKWRvanVsY3VNIVYjQmsuPkVbdFVNMFVOImZFX09NZTEzbkw5 Py85aXNkLFlbSUJFV049Y1VPR2YwJ0NuMgpjOC9waGkzb2NHKiJWTmBYclVENUdPOGNK KFs2YTZiaUwkM2AnbTtzKGQ7NFZEPkgibCo7ZE5LZixnQClIZ0dfYgpSWjgjTFxIdHRA LHJEazUqUTBpcUVBWVVmUjBtRF9FUWNKQ2cjQ2wjSldvZ2VvNl1zJmxqY2QvY0plRWBy TjxRWAorVFdJQ0pAWyFLTEljcmpRTjUnZSIlU2xfYGhUVV5ESD5kYj1jTCVwTmlYXFdY c3JpUFZTbFolIzUuXVhlK3IwUgpuKzQpRG5oWmVAX3N1cEVcVm9QNW1IKzZVLmY2QG5m Tzc2J0BLcSJEJktHN0w1W0slVkdUPjIqLU81USJqOVFgRwpGSF4zNT1EU0BGP25EXC1s U1g4Tk9xa2omZGhhTmdpKjJcMG1rMWVuUGAub2Q9P0U2UCZAZzo0VEZRbXU5N1ZxXQoh bHM7bChHJEtxNkk0PU5GRlkvWW0hJytHRlVTIi9qPC81XCRZKyRvcWpBVXFJdUdKLUgz VVduYTwzV1kjTjVbZgpaKjdeTFpkTClTKj5zayxFPCpuSiJBQjRxVF1iWiwqJD43cUk8 ZjUvNFI9WidZUjojSWV0QjIjO3BcaDo9c2I0YQpuMCcxT1MvcSlkU0RdXEhaRG0rblkz biswXWpRVzpAQzw5NGc/SkEkaFAwIzwuU01oP2ktOGU8OSs6NSRBKUkhNgoxJUFjLmE2 KFMtOkBSLTxhSTQzLT1JW1crYlpBKU5gSUU9WkZePVlbXEI1aU9Zbj9OOTkwPlwpXE9y WDNrQSJaVQoiQVhvVnI7UkQ/SlVibyYtbG07bFxLJUgkMyhkUHQ7L3IsJGVEVmVuRUtu JFMtL2dTWyQyTCcpYEdBXlRaUWZlJApRRDkpZW1STjs8aVcqcldjbEoyMjowYkJqcG1A OT8jaypnNjhIU2pUJWpzXCopPTtLY0leQlZHXW4+NGdmPS0pRApLaGxuUTczKS09LFBg SWJJLExkaiZWcTVMaFs6bmJHKS4ydUxsUGM4IzZ1XzBOT29QQFhYdVpoPSE1OktmRTMl PQpyJTxPR0koKDI5L3FDUypKVU9hWnBFUShGK1UhX1ZpPGs+PFBWZ15XWmBgcnVRO1tt cCFmLFtoVXFyTDdYQXUpJgpBNFpsc1YzLk1LOUc2RTlacmJrPmdJO25VUHNKX0UiWDVh ZkFVdXJcL2pLSXNcLFpQRytKOi9xYEJZVlkoJ0ZPPAo1SFM7YD9OS0w4YisyI0k6dV1W TF0zKCYpVnFsT2E0a0s2QkBRVC1mY3ArNzUqPjBEPjYjQilVW3I5dDVqbVU9bgouPUtY I1A7cHI1Li4+IWpSIzhpNUQ0YyFyVGZNUHBdSVVWPyltaDdbPDMzRFMkWzIkLSMsVi9Z VzdxVDwrOjVDPQozZUY3TFdHZWgmV1wkYm82KDVXS2tkbyQiXStWZHRubmhTR109J19h PiEqKSMnR1IiU2ozMGJWZWxZMS5mQDJyRQpbUHFKcHBHXFdPbE9fZDVPO0xNcyRSdGlk Q0I7IzlBKT8mXj9wRDRoPnIuIWhdRFYlbVVBRWJVTWkrQktiTzQkNApnRGksITVYU2ck WG5yPSlfXiNWTUNGTjMva2I0OC8oQSlwPWxfNCVPVz4wc21sIXVUJiVBZ25hTFZLOzc8 QCpEOwovRlBWZi07Siw7YEU4cjw3PDhKIydQZHVYVFJvclRLXitDKnFVQVc9ZV8yTnBA Ni1RWS9cPm9hMlEzVFBoX3JBNgojX1NoOz4mTGI5KkJIKlxxTSxqTnEzLVJhLFJqa0FU RENKR10uRWZVIXBiNWdORW9XIUpPa1Q9NSEpSUFDdWMwZgpbR1hcVlBJZHEvPmxXZkUo RyMsPVRQQzROTE43bistVitrQ20jOmJmYylPNXVtc1E3XjZxbUlhW111cTE0VShgNwoh Sl5vMHJyKWhlYSskOiFePWpcMWUjQyU4VEdaZ2dFYzpsN191dVImQ1BlOUwjKGgwQzVZ bE1YWVkta2xmSk8lVwpGbGJTI24nb2Y4IU88VGxANHAjO0htbylWIjomUyorX1I1ZSNL W15McU01U1FZRTBwSSlqSlU8T0BHKSJGTVNFUApkITpXXzslLDs0KmwmcztlI3RnPkpZ JEtJI2NKamdyQFBaalguSygsTiZdazk2OVFLW0k3QXNbOSdEP2trQSR1NApcK0piQ2tg WHJCSU1sJGkyLls4TUBsIj4wPlImNjcoYF08ND8vYGdgTydrVls8PzUhWUpLTWRfY1tx TWktXTNLPwpxOUM9RGopJ3QpbGx0TVMvNmQ7LF1RRTpRODJVJGpaSDorcUxyMUlqTig1 WGk3bCVbJmxgbFI5KUQ7PisqaTFcKwoocGdkJiVQPj1ORnR0IzE9aSpvIjppJ1QycnFj SU5QYGEzPSxCRzxvXDdqMnNrOF1uXF9zVE9qP2pBNzxhTT4mVApYZy0mYkxlRm1dUlFW ZnFiVFUmKDddcitPWTFPQXRVKDReSCg+WFJlakdMTmhhOipQSkxWbmwxTiIlTS4lRkp0 bQoubD4vUUYpRVglQW84dStFTVpYSFNbZnJmJG8ka2Uxbi5zQlZMQyleMio0QV9vZ0NW X1NoaCRxYTJEIylPLy1kdQpHcTdkLnMtJW9wZEBLSExeWjQiO1ZHbS04YC9lNW4zbWcj QVQlSDBhQHI4XWI4WCZNImJlZy8lLytdIiNEWFVyWApvSV1FZywxb2JHYlM3bXIrcSZA YW5PUVFaKlhOVHBvU0AmdDJZTE4+Il44MiIvW2goXVspaUcyKVosV0ZDTl49RAo1dSZc NCcvLDNhcSxaamUhMV8hTiJpNFtSVVZiJ00/Rig8TUtNZDxAPltfbzZALThINTw3Vz9H WWNeZidnOk84NwojbVo/OCMiQEs7T1k7cTZiM0E8OEEiR25fXDFWckxIdGswUVdnXGhn UyZMNG43VFAuUyRoZWBHZ2UwTlshQGpRPgpRK0s6bWZxaFpjcGllL1pEJldYY1smKnNO PyE5YTgvUCFEbmZwaDBvcFkySUBCKy4kYEIyJTVLU0JXMiU/JjE6VwphVFFfKF0jWz8l SDdmSTBPYGZRWydvaFhGLCFVTUhwYiFdZ2ZrRHVtI01qS1ZEV1tbXy90bmMxaGMlVTQr QXRTMwokLCdIX1cuU2tAX19sUlY5dUpANCZxJC40YDQ4OS5mVDMlQV91VylvSlkwSTFh JVBAMm9MdCRkPGJWS0lROChNPApTN0ZGQ0BnRC9WaWBtLltFTSI4UCpmLUg7Ny9WUiVM OVZ1bCUkTU8/bmgtLzRQcUJhJT5pSGkpYCgyZkZQR0NCKwpvcFZHSSRodHU9SDxTOF5n Ji10WzliZW87IkBKQEkwW3NsXU90O2khUCRGWnFRYD0uc2E6RVlRckZQaCJASChsOApQ ImJXclomX1xvQWdkKnBTa05VUy1cI3Ayay1lNUZVNldzSjxObi03L1BqakpfNjxAdVg0 UV0rPlVuMmpuMVVrPwpjbjBdSE5JbjMmVXJbcmM3QGAscyppQERMUGdgcEVxRGlkOkE+ WXVXR04lLl1ZNFlfKSQ9OUQzMWEuZkwlNz5qSgpEIlFiZ0o/MiwwNzwmbSxPaHNYTEkh TylnNzoub3A+b3EvIjgpKVJiKFxdI0VESjdHM20qTUhqNDdvTVAkZVIuVApUOVs6LjRD SiJzKiY3NFMrYyVGVChvVUxqWmFsWyhxV1tsbV9CS1RBRW82c0puZFx1P01JTDhSWjdc YlYlWzJcUQpGXTUnRDw6PyVqPDM0NkYlbVBhJVdnK040MzhbXyw0XHJvMC1MTzQwMXVo OW4qXSJtcVRlOTJLa1BQIVpXTTI8JQpQMWo8XzJIRUNhUiFsWmpnVTtFMFUoOyRJM10s MlpFVzc8dHFgJyQ+Wm8/YCo4cEchbik7I1FzNlM0KDVbMTQnRwo2Ly5bOS1ZLTMxNl5o YnMnK2RoaFteKTo1WERgaDdQcjFWMiVHXVwhaS9PRiIyZ3NQWFUpQCxwXlBgUmg/JU1n SQpKaEoyaUJzYCo3KChAWDxBa0hkVUsnbic9SWxydDg9SUZWXztCZzIvL2Fbc0dCSUE9 cWAyS2gqTGFqRl1JVXJwOgpHRChvWlQ4c2QsSignZFIkM01jImVSQConSmJOcnNfMDte c2QuUlojM1RlcF8zXk9oTidOJFFFQEZVU25YX1AsPAoxIl5XSExZT2slVk89UjJOK14t bnMxbSM+J0hAbFosOWAtTUsyNmVWODVudTQkS1pdLihIV2tDPlZPcFo/VlBrSQpnTTlD PmgpSjAsZi5bb0hkPmohWjRtbmJlRidaSSJOMUJ0LjFIU0dbaFEqO3BaYGdQJ142KkYs a2ZMTnEqb0dlSApbaWlgIldPMlFvW11iUXM/XjM0Y08ybDhPaiQkISNhKHQxTEosIl5Z XVdNI0NyTTgkZyMkPTlGbTVjTCxga2M4YwpycnJJK2xuXnB+Pg1lbmRzdHJlYW0NZW5k b2JqDTcgMCBvYmoNPDwgDS9UeXBlIC9QYWdlIA0vUGFyZW50IDM5IDAgUiANL1Jlc291 cmNlcyA4IDAgUiANL0NvbnRlbnRzIDkgMCBSIA0vTWVkaWFCb3ggWyAwIDAgNTk1IDg0 MiBdIA0vQ3JvcEJveCBbIDAgMCA1OTUgODQyIF0gDS9Sb3RhdGUgMCANPj4gDWVuZG9i ag04IDAgb2JqDTw8IA0vUHJvY1NldCBbIC9QREYgL1RleHQgXSANL0ZvbnQgPDwgL0Yx IDM1IDAgUiAvRjMgNDkgMCBSIC9GNCAzNCAwIFIgPj4gDS9FeHRHU3RhdGUgPDwgL0dT MSA2NCAwIFIgPj4gDT4+IA1lbmRvYmoNOSAwIG9iag08PCAvTGVuZ3RoIDUxOTkgL0Zp bHRlciBbIC9BU0NJSTg1RGVjb2RlIC9GbGF0ZURlY29kZSBdID4+IA1zdHJlYW0NCjg7 Vl5xYWA/LSwnKV9uMTZjRTo9M18wTzEqS0tXPz86LGBySDokYzxgVGRPLig/bl8pVHM6 UWA1T2JNS1xeIzJECkMkKDB0UVJXUFMwNlcvKj11LlxwQWBoL3BSdT5II0pXTz9JXT8j WDFzImlWSE9VKSNuV0BnPDxeVVg0OjhXJXAkCmhqWFtzSV9CXjtgVEhcV3FRcTYyNyNF OGBFVyslJDRkaj0iZF0mW0Fjb08pcGkmJypebUswN0tSRi9DRU1sKlk0Cl5VXl4iRlQl I15jazE3aV9IYE4qLjdub2ZxSGYkUEghYCR1MXRBOjZUbzBsPWFtJD1jTWdWaWchUjVT IWhbR1JqCmN1cmxzR3RbM243QVlQTSdoVmEyZnVdLlkybWxhPjxcZWJBQjEkUCZnbU4+ JmBjIlRPUEEwRTFuVEQrWlMsbihECmwiS3UxTWRKQSxKWXBPZ1kiWXBnOE9lIklnMUVH MVQlJVM7XTQ3byw2RDsrRCsmN15SNycsSVo0IyIvOTQ+MTxLCjlEKSFxM0xLIUVnW0I4 a2wqI2E5VEQwUT5ILyVYY0NtP0RpWiJyJmNNN2gtV24oZ2htNzAqVHFWNVlDcFRRPm9N CkxUIzROVTxdZWpFWUUpWFtzP2VePjBtZGddUydNS3IhVTA+ZSdSUTkpPSs5IjlXdD0i RGdIXzdKIXNsWiJRIWYvCl1NRTRiJmdJSkpAYlZdNE9mZyw8SSRVVTJgIzdkOCgoZ25s IicwZmRhWjxgSlQhNCpBbT5IRydcbThbSGA3STs1CkEjVXNmO2gwbTdGZVhRKEktPWRK WEpBVDJSXWIqTTBdYidNSTF1YkApLFksaitqNCE8W25SJiIvSmg8ckpXay5iCjVeVytv azNEWkIzMmlNZXE7SXU9ZWFpQkVEIWwyZDhqZFo6O0ojJm1bKGQ4aGk/cy9ZU2Y/T1Vi YEo+dGo5KUBWCk41RmVCSmZdZXUjbSxBX2xGXUssaTpSMSstPkxNc29ZV0lgIzRjOnNY N3U9QTBicGF0J09ZLlYoakwvU1t0QVNLCmtcTiNGbDArKlJvMycmMTJWM2c+XG83Mj9i JlIiZ2k4TTQxIWc2X3MuO3A3UTM2bmFDSWk8LTQua2BYNkQ7cmhSCiU+bDRcRjJLMz1a WTg+RyctcFFoSzNJdWkrRUJoMG8xMU5vRSRdci9gciQxZjtXY25KcG0waSYhQ1I2dTpZ bTJmCmc1XFNIWi00O1lwJzwoUUNUXEViPmU2XzxwOkoqaFRHTi5ZaSw7PUc6XDpkUj00 J1E1aGBecWQ0cFFaPip0Wj4wCjEuK2FoRTkpQjtHYSk9TiIvRjdkQj1Qb2JhKmQmQSNU SmFCKVVOWDo2amYiaVxLaFAlXDlxTkhCNypeJTUlV2E1CmFudDI8Z2s1dUxtSiY3Mz1g KGhhODooaVYvTixkYFRPLFVOQyxXIycuaSVkaUV1QD8qXGVWW2BPN20wYkYxOjpMCl83 YlNNb3FXSk89YFI6VC5EPXRwU0xtYE9bJShqW1JGLURqZXN0K2RpRE9YWCJSSEpaJnIz KVJdO081WGZQVFxAClpdMnRrVkQkPFA7QXI7OGFvcDM8Smc6W1I5MUgjZ0Uyb0xGPFZQ OmxTMCpbNTllRVQ0PVpqLydUNDVaaUJBSDxNCm0qWzA/Q14pSEpScztCQSVmXCNtSjFR RkFFIjMlKVo3JSFqOmIkaG5WakchPC5JKGhJXylIOTliXVdgUV5wbjFGCkpja3Q7bmZV Ji47dG9bbyFWTGdQbkBNZ1tPQVRCZFQ6MEhBWlZmKXE8cFJCMHBSQEdgKi8+UnBRa2tP M08sajNqCj4nK11OaGpTXypgaWkqUGY9UE5gLThoW3IvTkRFclhtOF84bTFnY29wUVgw cG5qKEwwW01KViEkQlooWTA/PV5aCmlzcWVDQVtdYm0rVlIsJSpzLlUxJnRNbVpjXTI4 MywlTFxWQmBoYE4mMztXMVdbbkFeQ0ZDS1k7P1tvKjNKSkc7CmM2LjlxK2IlVVdMWW9t ZipgWGtyWU1OUWVsJGFsU29eIkdwRHVwUWxHLTM0VTBqcVk5KDhgQXFfdVYwVUVhPiEk CkgzLzl0NlUlSiotVjAhPm1oOm1ZaWNPNWpFKmg/OFhaayteRSpwSVVscE5kXEswRE5q OWQkR2ZxVHFCZyhgbnA5CkUjSztbWDhIdSRNTGoyJnA1OlNNW3FkRy04JW9wRzhjcmYh KjlVYVZOY1IxaEtLbnJSXXMiXCQqIi87SmhoXW8uCmFmLUc6Nj1wR01mMGQ3U0NiJWtu UTU5JWE+bio0NmJzPmRdOERtQENILkZgKzAyM3VCIjAydUNnK0pKRUJsYTNpCixnW2VA KmR0PjBPSWksJjxpZSRsR08zS3IjKC9HK1tGLio8QTo4XyhVY1FHUUpXODw2Wz9xaSRx amQhWig4TmgkCj1qITZYMy5KMEpvKTFPKGJTYlpgcGEjVy41Ul11OnE6Kz02R2soS11A Pm5UaSdHR01KW1JJSE9rX0pgPl9VKGdkCmRpJUwhU2RJdCU/aC9OSDlaOGc2VkxIQWY3 MGQ7bmAvTzc5IkYvJjBeVWlsPjpRazE7UyVNalxeVkBIaGknVDxkCkxGWyg4cEFnPisn SjRqWChdPy4taCEsa2VoJmxaSWZvZD1SUnM3TT5UKTxtSEAoKFYuZEFcJWohOlU2I01L KTwsCiNTSStOUEJCT01rJEJPWGtTWTdpYDpYQ2gvOmtWbVRYIlAhLT45I1FjaFAnTzsn LUBoMV47ZVtBTldBQGllQnJKClZwdEJnY2t1VGZWYHAkbz85NVpBQlBzJlk+dUxhcnIn P1FcZUc7OUQ9XVRBR0UhXHJLaW4uazxrayU6KFJUPDAnCjc0YCliSDJrXHFbdGtnKDsk SnIkVC5vSFBcXi9EcikoVmBeZGNhRj47WidaXVEnJmI6MDVaOTlGdXJecks1KzJFCj0m QVVRYz87TDVpZUY6QWtCQD5ESFkta3JBdEtUcUZPPCg4T1VPN0EnZTRHME1mNDE9LjlC YmFBVitKKDIzcSUxCkpfXiNaRTsrVEQxM0teR05PQSouXTtyQ1s+T3Q8c2tgOVFcOVkn IyNSYytRZS0pPk90MmZtSk9NKG1tIz5AITNYCmEjTiM0JD03Nz9UO008c2pQSXFKcmws Klk8IzZBZVlbTkJjI2IvYGVjVU9FRCRxdDRHZ05cOCNNI2w4PjRYI0A+CjUhRXRKOU5X dU5RQDFUSyxDYD4iVVFLcW5gLGNCIys4YF8+Vy4iZWwtNDFyVlJ0TThfWmdDVF1PQEJc ZThFU0hvCm9eKUgjJkprSWo7bTNYTSpgPDlWJmMqPEtuamIqQ0MvXSluWl1gRm5QN2FI UWpDKW8rXmxUKmcrUWE3NikhYVAtCjxaX01tZyVfaHVEdT5qLzNlJlA8VCtxZ2QmRV10 PFpJSGc6ODRTSjZXZS0zV2Y+ciJtYVlCOVJKLGpsI2YxOVg8CjkkQS0qW1gpOmg8Xjtj b2pCP3VRZThSJ0hLNW9nYXJLNmkxKDRucydwNyxEVUApZyFVSURfYTdhSEclUWUnUzt1 CmZCbVhZXC1LUzQhU05dTC1FcWMwWD8qPj5NUGREYFc2Q3FaZUlNYlxnXkgkNENWWkQo JSc+YDtYPjJKbi1TY2hpCjpQOVRqXCJZW3JOLkMyQF1OYTpLWEtWJ0tQR00vOThXamU6 IUpPP1tYWTsocTNvbHE4cmszb1IsRkByXm02QyJUCjdbaTpVZE5gYllwXmhTY2U/ZV1n JE0sLVZUNzgvOSpoTDFANmQnXEI0LFhmZygxVHNHLillUW06Ml50ajY2cTBgCi1icWhM YDNzPzYiKjhqbzkyUSpMJkkoLzlxW0soYyZjUjRcKj9QRy0vSlBzb1o4XmU3MDpwZyY4 PjQhayo0bGQoCl9vcmRackQoMF9ZcGM6UmhgdUVhUzRbYjRsLCREP1JkdVs8KjskLm1s LktUblJaKj5POUlAIy8uVXRPZkQ3SCgrCidTVjo9YGhoQCFUOlBuJW5SaEI+SjBcJEc7 KVtEWS02STU2XihlWSVlUGMkcXIyJD9cP0VDZGMpakZcNHJmJ2BHCi1vKmYkREROXlku ZXJnQChkVT5eUkJYW2Qyczo3QV9DMSY/cGxLXmBIYERNOGRJZ1tDV1tJRiYvMUBUbGNK Xyc2ClpRW2EpLjFlU1E7TExvR1dEXzdHIVVaTypSVHA3LlVLN1hEZXMrOnVYQzQoSS8n WmItU2MhUThAU0FebU5AJ25iCm9rWCpEcCJkOmBgOWwpanFIPHVUNlJPWyJyb042Sz1V bTZzbkxjR0ZtU3FSUlpzcSVRQ0xnTy0oVi4tQ1laQ05gCmpqZ2ZvW0xrcEIrVEpaQVYl WlpUYWQmck0kcFFHP2lbQFZvPEtbRnUwTXArLyZMYGFCPCM7PSVkJExjaWNwLXBBCmM8 X3NJPDxfJTVubzByPTNPPEFENWZnVDlcPDJPV19oZGQvRW1vL2w0LkpQZVs3YFBlbWM8 UFwmZFtMVlwzbzUuCixETm4uWyUhbyg9S0dbWj9rY3RfMnIqOWRpPEdUQVw9cCdNaUtm aVBCV3NNUGZYK0lbcGJmJitYVCdeaWsoQ15UCi1rKFxdITtaZk0vT2ZDakxSJFhMbWww Vk1IJ1ozNzdxQlA8N3JWL3E2JiNtQSRtUDBnO1xqNCNsXDQtOW1BW05fClM/c2s6Y2hZ M0ZGc1UtcW5rP0VZW05AZzFkaDFaNisnbydyVj9SNS0yPDJFPkVDXDh1OiFMMktCIlkh UWdmNi46Cjtfbk02O0M2WmUtVWJAL29TS0BkX2xVcio7VXQwV2IocUYlXWxtQk1bMVNn X1cwdUBJSyNVKEM1OWArNyxZP0VBCiNOJkdYIUs4SGBdTmhZJGNvPVRjM2BDckc+MUcm XW0hL11BV0RqcVdZZihuKlpJSExwNnBRaT4sJXNNOEJMaWYxCjBoSSQlNzFCcUhlM210 SEIwb0JSQmpsM1E2YFleTGxnI1gkSEkyVURyU25uZ3MhVVxTVVc5RVRyOm1eYyVwSGVn Cm9cNFVnaj9bUThDVXE3VmRqYEAvIkc+Kjg1az51K2twVEpcOS81bF1bPjE7I1ZAQVU1 SV8tJ1RJOSRsYmctWTQ8CjYudEtkSWFRbFVMXT1jaklfOUIzVyReYiYvTypvW0dqZGhn VyNKKzhtZWIxTzxDNFVvIShZbmc0aEdiNl8oSHQ5CnF0KCYyKk9VYjBKJj87cWgkSDMw NGhkMUUvb08paG1CQVFhLWgmb2NYMkVjTko/UGUpW1FEUkokJ29wZT5bUyJMCl8pTW9B YVIlZWc0TU9qQV1VWk4hOT1VNnVPSG1INkIiaktXSTIxREBDPWZNblgwRmFbJXJMREZg azlCJGxnYzlVCkdvdUdUakcjUmpuKyM7bTRUQjJGcmA+UCxkZHVoLDIjQ09NI0ljPVRI MUZHN0ZpaGohSUdYJUo9ayE1S2twPFw0Cm9pXzVkQW5DWDxmZDN1RW9pWWVUY3JlIUw+ JG1IXlJMRj9bP1hjODNGMWNcNz5SZE5zKWI+Z3JfLC8+cjp0LmE9CjVwYlh0UGUtMiVE JV8mb01eU09nVjdGJXA9V1kmWG9KSy4xYDdFIkNzNVo3XmxoUEBeX147YHFJJzo9Ll48 IVF0ClZpdWVnMyR0bT48S19VI1UqNTRfNlRTMGJQRyw6VFcuRGc7M2BJJEk+Zms2WDYx IS9jNFNFQl5FVEBeLFFsXXFGClowaXBoJlUoPWI6WE5xU3JfP2RyU2NqZ104dCI4XHBO Syk4US5WdEEkIWchLkQnNCdfcWZRMCFYLSd0b1ZtKSRWCi1vKCg0ODpeMDBCMkRaQyJH I2o6L11xVy5EJXUmX1svN3I8QENsVjE/JVJURUdFbHRWOjQ4OkFiJHMxa0A/KlleClVh S1lRLS9hcCMvVSlQXGF0MmBdcHFlLV9ESFV0bnJTWnRpamduXiNNb0hYY1ByaCZpSz87 LS1aKHVpUkBfOm5FCi04QU1RaXNUKjVXZy8obCc5SShVSWdoMXNuTGxVUzBPUS4qRUEr VzNrTSNXWDVGW1AyUXFzXCRAXGxMWkVkMGNkCkpVMEpLKTYnYUo/TGtWZis4US9rcy5q JztQMHNMYjMwUUYrWy9IbTlIRmomUl5nPjlUZk5TJjBVOjFyKyJtYl5zCiFsIyliM0wm SitYLkoiUj11bFpsbkdmS0NpUVk7Pi9gZ2dKYFxzOSZcJVVrcFdfbUohJCFkQkQ3JmZg bEkjLlhrClsiUiU0LycnIipxKGFVaFErTkQrZSRLQWowUkkpQ1UlPy4mIUlKYzIjNk0w I0RxOiI6SXBrW2MlTHJQPVk4T1YqCm5SS2tYJ1QtMj5EKlNZIV1RU3MpJUFhVVM4LXUh KzpeQSg8ZTNJRy4tJWxKPWFCO24vazNYLF9mdTBfU0huNlVPCmRiMGs1LjQ0KE5gTWc5 Z3JwaGIxWjdKUGVqMiRPXyIjXyVQTnM4TWsiTyoodCQhPzlYUDpsJmRJTnBAPiddN2BL CjJ1QFEkNS06aShdQGczQ0BEO3E3PVNwcGRHIWlqKSY4JXFzaT1jIkdWSVcpZzsnSGQ+ bmJmb01ZMC9zVjlpMWlhCjs2PDFZWjM4UF1aLWUzdERndCM9QDJnbD9tOEVBM11rbGlt ZHBbcjZGLi8+VVpKVVE/WmhHbHMiRU9GRWhuVmxNCkI8XSNHP0hyU1FWMXUkJ2Q+X11a TU1dTT8lPmZLQkMlYUQyZnNkV1FKXG0jLUkxPks2b0FuLWBMOVNhODw7N3E2Ck87X2RA Xi1RdWFYS1cvOG5HPztHWycuQDg/QTZrQCRHVWY4ZC5HU3QhcyEsVU8sT34+DWVuZHN0 cmVhbQ1lbmRvYmoNMTAgMCBvYmoNPDwgDS9UeXBlIC9QYWdlIA0vUGFyZW50IDM5IDAg UiANL1Jlc291cmNlcyAxMSAwIFIgDS9Db250ZW50cyAxMiAwIFIgDS9NZWRpYUJveCBb IDAgMCA1OTUgODQyIF0gDS9Dcm9wQm94IFsgMCAwIDU5NSA4NDIgXSANL1JvdGF0ZSAw IA0+PiANZW5kb2JqDTExIDAgb2JqDTw8IA0vUHJvY1NldCBbIC9QREYgL1RleHQgXSAN L0ZvbnQgPDwgL0YxIDM1IDAgUiAvRjMgNDkgMCBSIC9GNSAzNiAwIFIgL0Y2IDM3IDAg UiA+PiANL0V4dEdTdGF0ZSA8PCAvR1MxIDY0IDAgUiA+PiANPj4gDWVuZG9iag0xMiAw IG9iag08PCAvTGVuZ3RoIDUwNDIgL0ZpbHRlciBbIC9BU0NJSTg1RGVjb2RlIC9GbGF0 ZURlY29kZSBdID4+IA1zdHJlYW0NCjg7V2o8PkI4SVgoMHVMYkozUVtaTSUsT0wkKVpG VVAmOG42PCVQWDUtLj1bO1h0RTI8Kk10NVdpP04+Y08+JDkmCkRkWlUlckIyc25jV24q Z09LdERLLWwhK2sjRU5wWm1MLiksaEBGNj8sJXVHV01cKFZtJlRUK1I0OmZcPDRjSmZV CjVGZyUqXGhASlM/Z3A6UEhbUnJCSXQyKXUvZlBEYjBaRzxUTGRVKlFrLDVtKUctRSZC XjgtNyVGYFRSN2NkOyVJCmUrcmUlTWtBaiJBcGpQKjVVNTk1Kl48NExuQVUiRGVLczJz OVRHLidfbyVdLkhvKV44KlBNZ04yNkI1Wi1rRF5oCkhkMkJZIzFmOz0vci5uYzUrOCUj am10aixIMlsqUmxkc0QqYkdTcllOYkBhSU5WTmhxcDRHI2JcTD03VT1PPWwkCiRhb2NO TWhNJlNmIzksQThsVThdWDghazo9PTVDdVhSZHBgajFJXzM2KDZcZik8ajFQPWRuNyJC aid0Ly4maS47ClppcixtYmZ1QURuVlFgKVExInM5V11EVFUnIjVRJWlsTSI2XitKQEA2 KVpOK19cVkcpcnRYYmUhRmElQWQ6XVcrCi0yUV8pTjBaVlk7VWMnVUxPYTdDN09wWCFi UiJCOUpJKDlBNEZvakM/YGxlSTYzRmMsKVsiYXQ9VS9JQVVvclNWClxfTjpvSm1mbkNh OktYKUM4T0pPM2F1ZUhyN1BhLDdCKHVxWmthUG49cjw/RDVcPDlgS2EpMidqMTEhbSpK K0VACi4jOSk4Ni4zck9JQSk+Pi1lJDJecHMlLC9WRUJXM3FGIl4nKS1DTidMUEBjLVQk PCImZjZdaDcmPD47VTwiWW0rCj8hY2gzSFRYTiNKYF9ya045RWk+WUpBXEE2U0ZZJFM1 SzBWXVczLl5SWDVEZF5VXWRVVUEhK2M7LVpmdFdOMF9lCklQaG5eM1pwNU9cc2BoJl4q MVRUSm88XmZiVl9kVV5aQWk5ZG1AX14hRm89QS43dTBLVXBcUjpkKSg9NyIjOXNuCi9N c3VYXy1EcCM4ZkoiNkdDPWRXMTtDPmY6PCsvMCc2SDkkYTVgVDFmJHM2S1pmPjI5JFdx VS9Hcy5gLDAwKSdTCklMRC85VEEwPzBwdCY+TVA4ZVlcKywtLkk1cV5AX0IuSy8laGJf VFkyV2AhIiVlcFRoPWZNYnQnN05tR0lxaGNVCjooVy9PVCVcWiVgdCFiVU5DYC4qUWon cCI5Oj9aU2pxdVlVUlVcX1ZpaUdTdHFfIj5vcFg+QCpxcj8uLGhra1pWCidmO2cjT1lm TiElUTJoc2tzITloKzc9VEEtTyNiOjdlQlk+O1VPODlsRnM7XkxqNjVaKzMuckVaSyow N2AqJVxYCjI+V09PbWg1OU0sMEFnby4zPiUnIm1hSGhuUks2ME89W0YoOitdKiwqZCgk aTNAMz0hOnFnOSMoJGtkOyNjPlpVCjslKktbYCdxPTRtb0tnUTgxMnFhZjBoPVZoWlBB U1hXV24nJHJSUSErT3VoYjs/TzM3YTosIyI8LmQ1KSNWcUJpCj5IKj5pNl4zOGs8JE8o S0JUUiExQHA5PGk4aERscVtFdDtmX0xTXWkxJzs4QzFxOWBUMGdtNnU7QFJuLGVYVEJI CihYcEx1Ny0sQHBMLW0/R0dgYzgvY1xTTlNuKiZ0Vy5LVF9zNnMkRE9TWCkkR2xCQSRM L15saSgoWjswY25lI2BpCkZUSihGM1ImVD1aWlZpZWZvSCkvV3JnWWhDRmk4VjxnJE9Z WV4/TyViO1U5SV1FUFNOYTosIXVoZ2siV1dFak5xCjJIJy9WOzxHPWI3cSxaSytQLVw3 W01EYyo+JWdIQlRlVzZwIUVmWTwjMVFnaTsidFBUJj07SyZhR1JXdWQ8Rl0vCl02Xyhh KEZjQSpQV1U0bE5EKDQ2bFsyUWRxY1AzVm4ibVwyMTReS2tNLG5EUyFOUDRTOSRrPUAx VFpkLk0+Z0o3CjZKRiszJmsmR0dhbzoqKixNQmlMXl9ML0tpU2JyazNdVkosRjJScWtc YGE0RkJtSFNILDgtbEwqay5bWSc0b2AjCmBsXEBWa2QtUEFKUzYxa2RNQmphQiQzWVdw Njs6ZU5xOU5pQFZSU0tCXms4QmptKCNdLHIwZllSMFVhbl0lRllrCjVpSiRiMylkPWhn bz1cXzgicE5KYD1MSy1lR1wyI0lcNiJEOHBsREtUclpjNjNuVzA2S3U8dCZiMi0iMVgj XzFnCk5vdSJISitsLi9gTkJqZFdZVUwkViE/Yi1YRjhQXTByQi9pW2NyM0xSOWtPQzol Jl1CTVFHc3RRZCo6cEVEamFqCjRSVktkPTEvcS9UUS5WcyhbMC0wZ09BYjNEcHUlRmdW TkFpNEdWSmcjNFIjUm1ERCIlTEBCWiptRWEoNWY8ODJPCjFocTlmVypwYEBlNmxKLGk2 NiddMy4nTiIuKiNjRTwtcyY1X3IoIyYrKyFDamw4KjtsV0FWI0dYNkY+RjpYQVIhCnAq bjs3YEk7PVBgZFFQZitkIkpjJWJpPFJPXyMuN21bKy47RFhnKVNqLkxFdVFfUW1HJlQ5 UUYxYyJPKlVqWS41CmMuUUtecD1ZaCRaUSQscVthLS5fYG1lMyQ3YXUnNVFASl89N2gn cDxeVHAwLDQrOV9iRVZdUWY3LipUNCozKCVaCjU4V0QkMWVSWEJCVSEsOVY2bjFKVFlW TGJuPV9IM04tXGFqTz8rITYiY1thckg3JVFwbDtJQjpAOkY5NTlpPiRJCl9nJk1dWSFv LiUzKEM0XiUrRVI9MU0pLVY8LmQ5YScvXUMsalloSz0+J1BoWXBhcT5rNSxhc3E+VkVE XTQxKDBrCkdEL2xPa2JwVGE6TnBoKkJlbzY8RWlwPzgqSl5QYEQ0a1FmQEZYIy5bMj1L UFVtO1hiJGxXZ0o+MSk7YV4rayxxCltgb29qUU1FREpAdCUkTUZFZ24mR3FpLjMwY0lf bmI6U0NlLUdFOVcsKi06M0JoLT1QSWMnXGFeYmM1LzAmUjE7ClgxQzdKJWRnLFBfK11Q KkFrYFIxViwnI3FUOEguPlg3LzJWS2w7L01JXSc0UEZzRzdSJ3UmIWZWI0MpSzg6SnFN CjYhUklwb09wZyFgYkVRLF0vJHFbRDRUbkMrRDQkKDkoKS0qSlFRajYrXjRELSdkUVxh PythT3IwUTJjXTVXMmhgCiQzN0JBbTRfUydwOlpKNUcqLVtYKlF1JyRbaywjUC9fMSll aGEkKUBQQSs5czpKKCk6XCIwOFQ5RkNGVCYxcUE9ClNqSUtKSGRbI0RnK0g9Rl8tYlQ8 PlZGc0M5N1pKNlNEPVpmY2ZYVF5qJVIkUm4pKmohVS5RNSlFXVpOOGNtWTJvCktzQ2Vc S25MZUk6WmhKSENiI1xbRD8iKzokNi0xLE1EczxgMGUlbFBHWjNGcW1vbz9UTFZAUVEt aTwsZj8nQFguCkluYltRTVgpMSNNVCokKlpcbW9oRSM7Q3NPMF5MOUdUMTBKNDpfZERf ZWs6dUBfMmNTVFg6S0xtSClORy89X0BbClNtNyxILVo6LVhQO2VsW2JMLXVIRjxuO1Rc L2NeZ0gidCo9NFFgLWJXPUVnaV0rYz1BVl9MOzspWkBHW1RdJ28hCiYoTSQ7PExUY0c8 a19vM0ZicDtJIyQlUyloRlM4L24zY0Rhckc6RSdObyxvWlFpJmAtTWU7JWFMZlktYjdE YVAuCmluQGBMN19sWkRiOmIrUiIyb3BpSyReTWQ1ZUA+NnJuUUwmLG5CTi9jVjIyITtw JUtmT05PQFozKSpfU0VcSWhACmVrYTZQLEw6RVopQTFNMD9yLGxxKihFPUsuZ0A4QEsp VidjRUo/ZltWM0stKzkpYVhcJVFxVVJCKldSTkldTzNeCjVRRUJhTWdzWSZyaiFYKylS dGYjL2NHZSRSNVJzMEVIc0ZWUiU9XXVPK3BCNj9JZlpVP29OPGJFI1RJLkA7UmcrCkBs JzIyaEMyP1pIZ3M8ISxEVUBgWGc0Om8+L2xBdTtcNGBCaDRTLF9LaSEjRXBzInNGOGQ5 ci5QVCw0KFRvPUlXCkRDLDpJOGtaY15AcWhYMUFWSVdGZjhFVHImKi9CMXJPTiFmaGhw LTsrOU9sRzNzK0xlUGdGTVg2LkAsc2hQSk5TCk5JdWk7LStQXVdnZEdpL1BEJDczaE09 ailCLjUpcjUwXCJLaCRvKCtxTVFXIjtcImwkXm5zZUkjOSFgRlZkND1tCkpuRE9tN0p1 OS0hNmk3IV5QZmlXbk9rKSpVN3M5YEE5OCdZbixKM0lbY1BNYytTbDk7TE8kM2JlXjgs REZFTFMpCkkkKGBeTTZrakhWQCNYZThNRVQzYDZSJ2snSF1YOSMhUmwuVTNnZEhPWCZM MDk5R1E0WChDUD5XO0UrNl5RRlYqCixKT0VSXyprJklIITZnTTs+OUh1WzNHI2A6XjFt P1RGciwrZilSb2ZQcTRbaDVIYGImWzlhbTRrWWI8UldLakdACitzOT09KD0yWjkhYis1 XlNTc2EoSVxkPVElc1lMMEhzR040VzpmW3FaayVBbVhaOUhEYDghUl45Z0BJM107dDhq CmpwPW9hV1RFaU8vJScyQ2gqKyNycFM1X1ByaGhNVk9WMiRXaCwtSUcpdVclNk1hRGIu Q1Uva1RnU0wzNzNaW1ZXCjwoQWVNW09hITwyZmxwY2RZS0hhMio9JUFjN19aXjk2UF4s UGo/MydsbEUuaDRGL3NQOlprNzcyMSw6PnBuZzlECjxnVlZaJVxfQSkxSFAwPW1iUm9f SF9FMmMuPT5cXFRpKSldMTxTTnFMQE9wXT8vUmtFIkRWRjVyQ2RnJ05RWUk4Clsya0c/ Nl9KS1kmMElUKFojYyhvT0g9R2s1TDROYSYqcEBjbTEqOyRYL19fc2FvKE86WEdoKTlj OSpzXlZHND44Cm9wKjdyX25OVCwjWypia0xjVio2I2dIaTBJRyFAbjImVUJgOUxudXNa cytlciwlajZrM1U3Zmc4XmE7dD81VzZLClU6TkVQWSIkISZiREknTDBDIW5BbUtaV09K Yy1sQ2chUzM0TmltTjNkRUBgZnBwNDpFSzAoL3RGSzsxRT88Nm9zCixQbldPXFJrLTs5 WyQ2YU9YJXItWE9kK1NsJlFeOTdQQkx0TlJLQV4yRCotJSg3MlRtQ0RKPiIvPSZVZTwj bXNICjg+MDE+KDQ1SUE4KlVURFY8LHFORk8yJU8uQmtDTCRmbXFDQlFKQj1vRDpHXzpz LzRlUSwlLSU6Ii0rPiZzKW9ICk9hSEIoPCc3Q3I9WCI2VDJrRihvKTdYJTs/LmdaSTFY bS1DYzBqN0s2JWosbD9gSXUjaVNicmwoXzVITEIlOyEqCmlAZk1ZWS9KK3VNRUgpR0pl SShTJV1kXz8zWVNOckYiLFJcL25fYWU1TEtgJ1I2JlJXZjRAUCNPdWQ0WFJyIk51CjYm LTtHaCIxNCRpUSVlS0EnPTNzO2NdYTgicStzP29mI0A8OS4mPCdwMnRVNmtbR09bPC8q cURjMz8yO0xHWy08Cm4iUlAvUjVjXyhobCRVZyU4QiUiSjBdRmFMLD5rYSxSay8zQjds dWhLP1RidGMyNGg/QTtFZjdxZWd1JyVlSnVRCmlqLj4lY2trJnFjcWc9LDY5ZyknTE9q dSxLTCNabCdGI0VgXFVeQixAMDQ9XEwpR3EwTiEsKGJOIlF1KDBaa0xAClQ2VDoqIis5 KlxRWWtZTGYpNCM0TElzZlhUaDp0TVR1cGNlQDRZZTtAQFkrbUtsSTtFTSI9JE84SEw3 Z1M3TSpuCi0zPk0jOU4oV2U6LSJnTzMwZUhPT0smSjdWJEBpby5LRCNKXmMmXy5SKDFk dG5KQE1SSSFAQGQvVjtDWmQ+OyVJCiJtNyRpaUcjMl9tXCxRcF1UY3NcN0JlMXRSJHFa cFI1XFhDcCJfQk4ybTdwNzFyX0dFWlUqQSVSZSFhKjlpTFNFCl10YlBfKUllVGk7M0Rg LExhaSRlOFhKNU08Lz4uKUkqUT47UFxMMmlRWjNIdSYmSVMlJUl1RkhcUy40cmxOUXVC ClJYbDdyX1E0IlQoLVknU1M+L24tLXExblQiMl5EIVhxb01ELiohcmRBZmshL0g2OlJz akQ7QVRVbFA1JmBzcC9lCkUnVW4tLl05QSgxVV5GLUkpVC5fUmZCcmlRPV1SLjBFJTY7 Z1wiZFlaIkdfOF0hbyleSFNYdWQtbytgYmpvbkBLCm5UYUUiUUx0UV49YTdlNUIqNFhZ NCQrWSoyQ1dJOSlvam4kbChVRSZUYlIyKlpWIipBb0s0W1paby80UThPS0dVClZ0TWFS Iz9CbmBNMTozUSJYIiRYMWMwUSpENkFTPWMnLWphRSNDXEVTWzNsYEU+RGc/VD9ePD9s NT5XQVpNImk3CktHKDUpMkBxR1FXJydWTWApMzdwNmQ/WS9LZFZYLlxIYj1EcmBiRUIq bDlsTkNuN211Z2BrQl5xU0lqXi1eRWg2ClhqVWMvZFExcmtaOjFNcmwuWTIzRkxuUGJc ZDI8SW9MLTkoZEhMTDg4RyQ2YTFSb0RzcSpATWNJJzxvNTRrUF5FCj8mTyJNbVwnJXA0 YTAhX0NCMiFxYFAyfj4NZW5kc3RyZWFtDWVuZG9iag0xMyAwIG9iag08PCANL1R5cGUg L1BhZ2UgDS9QYXJlbnQgMzkgMCBSIA0vUmVzb3VyY2VzIDE0IDAgUiANL0NvbnRlbnRz IDE1IDAgUiANL01lZGlhQm94IFsgMCAwIDU5NSA4NDIgXSANL0Nyb3BCb3ggWyAwIDAg NTk1IDg0MiBdIA0vUm90YXRlIDAgDT4+IA1lbmRvYmoNMTQgMCBvYmoNPDwgDS9Qcm9j U2V0IFsgL1BERiAvVGV4dCBdIA0vRm9udCA8PCAvRjEgMzUgMCBSIC9GMiA0NiAwIFIg L0YzIDQ5IDAgUiAvRjQgMzQgMCBSID4+IA0vRXh0R1N0YXRlIDw8IC9HUzEgNjQgMCBS ID4+IA0+PiANZW5kb2JqDTE1IDAgb2JqDTw8IC9MZW5ndGggNTYwMCAvRmlsdGVyIFsg L0FTQ0lJODVEZWNvZGUgL0ZsYXRlRGVjb2RlIF0gPj4gDXN0cmVhbQ0KODtXajxoL1x1 VSg0SD11NmVDKjFENydYRGRnYkFJWCFDcEMuZS08YztSMDRZVCVZQG9nSXJiZSJGKV1S SGE+WjMKNE4ibnRUKU9AM1E/U2lZWigiK3UzRkFFTXAlYicwQ0BwWVksKGBraVVEUzpH MDFMRTcmPmEnSikwdEZTb11CVVMKY3BKOG9YJlQjbUlIOW5mbDhOc0lXN05BRmkncFFY VUJvKU1kRFUrPitHZHRiYz5ycENlIiY5QEdqPTw8YFI9Zl8KKTRrazIsQyZBMik5Jm9e Y1BXNm5IcUpkNjkoP1ZHOkRsQFhwbCssZSsubERYKD5xISdjVCVMMDRTU0I5aWQkKDsK JGshQDQnYGJncWhmalZKMnJxTXVKOjojLDZXYEs8P1RrYjsmWDklJk8nVWI/MEkxPUFZ JGhTci5fcGA3KmhcOmYKOnFqKlchbUgrTnEucWA0bSo7Rl5IQ25ydGxiYkc/ZVwxLUtu IVY7WkJYWzF1ITFsUj9NXD0qUTE3P0JJYTYqU3QKcVxZbCZIbGgzZUw8aGxvSkJWOnFI a0c5RUwrYk9jKWRgO2EmYDxpMW4qa1tVaHUhMTNgVUFHa1opX3FYUCRmRkEKVSoyakAs O2JWLU9sbHAySmVYN0s0clV0TydZbmUuWkF1IjIrRClYWSc9ajRaWERUZmMoQidUPU1y ZmA1cTtVT18KV05namgmXkN1RiZPVl0kZDRRQzoiLXJINitKKTZkR1MnSzhHI1w6SSQ3 ZjlGVF5WSHM/ZmJFQS4jUFAqRFpYU1oKRyVsczUnTjcxaFooKlYpUj0oO181T00pI3Ff SGAmW0cqbGdmZ0E4WURTSj1LaWRuRjMxR1k1PmNFI2JKQmtocFMKVy8jaG4vcjJfOyI6 XV5iSjg5WzljNUZNbjUhcXNlSmJEL0tjRnNDVGNbMUhuKGRhSS5gXDxWJitGN0oyWCRV KVMKbk0+KkgjLFA+Mzk/SUgoUzZVNl00RickLDVaailgNDI8ISVOOF9JI0pdQU4xWzNJ a0FoRGwvPEc1ZmZgLVswKDUKciZCPkNxLlUmIlZwY0FjXzhSLkpNXCE6QUVWMEA2bGUx WnJaYjFnaCYrZVE7U085WF1Da0kndDpgZSgvOnNUM3MKOWcvWS9rY0tFIW4ybz00XU5z QiFQIU5kZW45PnJGbjAzPl46WCdjcDlLbihGWjQ8Sm8vQkRzPjwqKnAiZmghJSYKRyRG JEFvIXVocllwN3VDTDJKYyYvYSYoJC1ELzk+NiFxT2ZQcnFDPWJVODEiaWg+OClpU0ct KlMyNW5mKi1QZjoKNE9eb0VEXVFbPEJTU1Y/I0cic2FbLys8UFVlZWtANXFMXUNtPEhO VC4zR0hYXUM3PmBvMmczZGhNJlQlTjk7ciMKPzlFM1VKWXRWMS1GZExjUSxmN2BwW19I QzpFUW5XWFlXcUgsXDltTzsuZ15sRDsoOmhdbl4/P1BEOVJIXSVJaiUKS0FjYy47V1I/ OltdOForLWVHVSJGNUVOS3BqcDVcM0ZoW180J1ZDQC5fQmA3ajkvPywrMmxPVmJtSCYq Ok4qRD4KNTdULFIuMWA/ODBYZSlHaDtnQXQ6cDE2NTArQFJSTThlYz9SaWNTTVhDNUpO XGlFXmBWUixDTEcuTGNiJl8hW0sKbGpwTCVibytFSzVxPXNQJik2QkFEYXMmUUJzbiQ0 ZmFkXCVDI2llUU5HP2BVajk5OFloTFJNdHBnWzQzXHNnPzsKJSZJcyJNPWF1bSVzMUBl KTwzImBAXlpiODgsSHElUHFWaS05NCQ4JV1QcShdZkx0ZVAvOXRZa0xBYmgwREFWblsK b1JGMGEvOVBBK2xHQEBbQDUiU2ZxZEZWVVFGYTomRl5fX2lZYTxqbmckMyViN0M5Ji4p KnM6M1IqTm5nKi1OR2EKYlNEMUE+XTxnJlZaI0A0WjQnQkZAUCdTOlxoMWsnSStzUFsz P2VaSERcLD8wJT06cWBqUSI6V2pxWmJLWCNyVjUKKUkkWTpbWSMtQkIoUFA1cCUuR09o KVcyK2M+VFZ0VFtKQGNPWURBcTpWJUwnSW4uSjI9Xm81WGAuWmZRXz9PSVkKKElNbm1G KEEyUTU9UXM7aSVsZXBYWjtWUls5OCU1OkMxLnNTNjgxXjpMXChDKipdJkg4WmR1ZGRS RmtqS3UibTQKN2NcRyEpR2NANDtzRmFFcjVKM0gnVltTVEdYP2YxbEk+VzJkdDJWRD5J PDNkVDhSV1c5XiRiLSdsQi4lWmVLJWMKPGpwXVZmQEBNNCxRVGw4MWBvJFotJ009YzBA bk5bTy9RYFktY3N1USFoXGQ1SSI8dCE7WmkzXlFxUUQ5VltGWl4KSENva2E3LWdyS2dM TE5vTjRNO15nLlpPR11DPSZfcDg4YiJvXl5uMTxTRiFrKEFeL1poMFgvKVNoVUNlLFBs MVUKPCJITlleVEtnOlg3N0hZSUFeJEhsSmJJVEtYcTE9ZD1rQWdrUUkoVFdabGxaZHAz SGVOJGs7PCNyYzxZZSgrQDkKbVNXN2lWaWUsLydSI2w1MidKTUg/ZnMpODs1ZyFCXS1N O19wJkdQVW0pXTlmLVcvZ1dNW0RjcycwNXA3SiRgS3EKXUA8VEIqYTM5QVxRNGlbOVQi WGpJPTRmQ10jZ1IiZTVVOktlWW4jVUY/Sk8/VUNbTUJUbkhFTUsqNWFnPyhRZUkKcCYp bDYkdFNkc3FlV2FwQj0zWE0rY0lvOyE/OFxFX2wtS0FgQ1NTVV9abnUtMC89WU49Xzop M0duIz9oWUFiSj0KS1ZMYDc/RHFRVWBEZWokbXMiUytkTU41KXIsbkplcSoiTSxfYmxp UkZUdGZZVG0za3BBbFonMU1iZiFFXkcvZjMKYyxyKiVKR1B1Ol1aZUBSPWZqMXE6clhh VjInVGM5SDVhWkdLNT02UVpIZmAhNF91Im5haUNdY1c8MiJnOyVOcGkKS1cmT3U+WV5J Tlo8RHBQY2lFNCVhN1heI0wwbDNeRjw1clcmXCgnUUBqNEVNRV9HJ1pnPFIsKTxLRVFe P1leVHQKWmg/dGRDczMlRDtWNE5DUkk/cGghSUtYJDpTSiY9UDtrdWBeOWtnOVphbSFi KScsczMsRVhEdWRzUEdaJS4zQW4KYmZWY0MxIjovbFQ3PilnKWteLGFeXWxmbF0iTTI+ QEZKJTNMKltLN1FuZ090XTc8QDZTYGxDU01AZFc2KHMvbVUKTj9UZW5nLlooSCQhdTgw ayhYXzM+NU5ROi5DSz4kN0J0dCtRLGxPdTtBMixfM0pBLG0kLyZ1cU9cdCZMWWFcb0YK cF10RWdIPW9TbDQyKzshSWQkPmdDR1p1ZEUjJ3JAR3EjLDs7MGpFamc0VjknXyVuSSFW NitiQUw4Wi5mVFhKU3MKQnMxJD9qTWFFZGcycnFbalFoKjtdJ09damhFRTxwaFhGP2Uv JTpYJSV1Ry5vLUE/byU9dD5hdVRIK1AyL1xrWV8KbTZvUVxRTHUxSlVJS08uXklob2Yv SUxTQ1AvVzAwLGNoWHRYVFtyPlFOU3I4OFwoZy4nO0NOKjpgTWNZPHApaU4KUUJfdUFS PSpBc0NoVmNuZS1qdXNYSHQ7NzpRUi9LRnI/QkFGO1hOXyM1ODY0N0ZLKypHTEZOIkM3 VlQtUDg0SF0KVVRYVjpYPFRENiVWJmYnXitVNmdhUFRySylEak5xRz4mLyQpYD1RJztD QHJuPU1FdVpFZCxVNFJYZUVoLSg8T08KbzgtSUNfJy9AM1dRTEx1N1lhZWJoPVcyZkhz Pj4uXTZDazxEKSlFaihlNj8xLFdEVmQmS2U7XlpPX3FkJkFcMXQKKEAoUXBqYFR1MG1U Q0lxNzQna0FJSUFaLUAyazRlUFZuWVNkOCopNE02ZyczMyE7S3VqPkBScCYnZUxlTiFx YlAKWkR0YXM7SHA3RUohLCM8bC5YJEJLLDg9clYxQjY5PixrOkpyTWU2SDtfPz9nNG0x V1IhXCZOUFFkdDUmUkEmRyMKRCc9SWxdPWNvYGpjMCc1RW1bVDE1MFApMGwvPjleT3Ii QkZYZkdcYFBUInMmV1w6Sl4/TyEuL2xIIllENUNlaSYKYEJhYltucWtnNioxZDZhWEsw KkxuRVJxJkI+YUVBLjRfRmMySC9eNVpRX1tlS0U0KS1wWmlTVTxrWz1lR3A5czIKZj8i ZykrOFJxJzMmamo2cydyXTZHMidJImtTTms1aEcrXGs0IVltczklYUE2TiNIWTZfYiwr LzphNTlcTz0/SEkKZEpROjNuZzdfcFVBclxgTFxFYlYvXDk9RUQ9SidaMyE0ZSVkcSQ3 M1ArTCFUUTcnL0EzZ1M2ZyElUCM0ZypeI2AKU0o+bEttcSVkUSVOLHFCPjklb1M1NHJa UEhUSFNmcDYtUlo5XVVWSzprXEtKa1A2W2AqcFdbOSJfSFxxPj1GZksKUkEwY0RAdTVB SEdoUTpPRUFUX25VLERnISdBVEgtYj1tTDBuV2olUkMqKiw/Wk1PLW4wLVVUJDVGcFJN T1JnbSsKLUA3MWIvMStcIiJpPGRiQ1kqX0lmVSEyZGkrUm1PI3BCTklLYjZFW1A4ODou PUAkZWNfSlJNaE1RNWVkKzw8KGYKXj4kNl5HW0BiaC8sMThKXi0qLUwhcGFsYCUtNkck MHU7NVAzX2loZ10mXGMsTyFmQUxjcEdAX2RySWh0Si5tQEEKR15LWGlGPTdBT2pFWlox Y3F0RU5XIW4sX2YrWy1TUUQoQCdBRFIxNiEnWj5XZDpPKSU4ZFltTWcnak5CXDtbPEAK OUMsUiljcCNZYWdUTGQ7OkQ5SnM4NTZxa2IqT1dUU0Z0RS8/M0RlNWtPXm5yL1ZIQVpQ N0dHUjcmPCEoP2w6PVgKRiEobzdJJGpjWjBjQk9Fa0hpTERrSHNTPScvMycyYWh0Y2xU T2onbEMrO0VESk08YkJyKSNgJzAuN1lDYFJAQm0KTi0vIzQlLnIlZT9UY0txTDUxRHVl XVwyUm1bcWdFYVZuOUVCLC5EV0FVam1dTW9rMnFGLzJaWFQ+O2wiUzQma1UKZnUzT2wj XXNLaDVdTXU2UV5OSkIzN1tfMmo7VEJbOWpaa0FZV0FPP2UqY0kuOnE0RS1hZUw5OEhV XE1SYUAzQ24KVWU8YDNPOkNtX2kuY2prN21kdTEtLSJxO1ZhbjtPXkZeWC1TWGtGM00n aj9GLiE5WCgvbGM5anMsaSp0RSpQM1sKO0xIbSxRaGp0TkYvPy5sXlJLN0ZcZXVCcldS cU4xMzQxbiI1K1NdUE5KamJxQCEuVyhKNi9lRFNNVVskSDZbSHEKWVh0cT5AP0cyUSpf ZWtIOkQ9K0BGKixxUC51aFYjKCtDUyRqaVpAITFyby1ucC49NnMrUEpZZmdTMWNpb1wz LjYKbGtnSz4oIjMiaWYsYCtpRilQIUkoWF5CKTpibj1xa2NYWE88Kz5CUDcuQmBuMWFQ dTVAMUleLCo/MlpoTF9LO0UKMDZzZUFAS0tpRFwtMGlWNWBKMD46YFBWZV1XNnRSbmVH ZGpvMkcyMUZEXjgmIjJpTmtGcDYwPUhOM1RLUnVcWDAKX2VHWlFAXlw3ZURtJDMoSj4z IzMhOXBMci9UIShRUU1PIyEiTl08Q0g8O0dYVUFJayZBMTJpRV9iPCo7S2hnPUIKK24z Ik1RcV5VK0BZcypcK0BxWC1HP0xwOURwYDEoK1czVSlXPikhUFpnVW9RcVNHLFRKPD1u KUJmSUNgXj1SYiEKZUtBJlY6TUcvODxDKl8iLjJHWTUvRFlPdThvRWYvTVtcWGpoTCUs SWtxZksvaC08NSVSaUlpKFtEbTxWME85UmAKISZWcjluLUhsXUwiNiltVChXZChxLXRY RihEPCthLj5AJVA7Q2kyVUM/SUI9Kl9dXT4lPUpHTlRsbiNIJTpLMDsKZkRzKUArXWFq Oj9SV1lQKU08LVZPQGRtQUpJRk1fOFFcWCUvRCg6QWFhYzlqY2pRLytzJ1VCOTpDMVBb UidxSFwKYlMpWmgya2oyKGwqTGx1PFBMX1slVmlUOEVJVC9oQEMoOVZJNTcoRVgwQWYi bUBlRD9UbyE8KWxtTUo3TSQpQEMKNGozcyY3TCY4LywsMnVfJmwrbCIuaUxlJyEvRCxm LVc5TVdAMEMpLE1QN1B1WjA5bWBkUTc1NjdwJ2AvOTVSa2oKZ0ElPWwqTylZdSRHTVtv K0lWZmtTQWpOVixSQnMzOGpfXyUuJ1djYV0laElHVVJFVClKWmVicl1uUWlqPFxiXCIK OEo0SSpzNW8rVSw2YFZPWD5cJmQmPmFgUGMtXmJBaylYKXQxP2t1JF09bG5RaC1Bazhx SmU8NEwyYTZYPG1EQXUKK2xnP1hUXTtQNj02TV9FL2FGaTNvRVRobyhBRygzLm5KKCMn X1BLYU50JWxuck9qNGFVTEEhImBTZkE5V0Izc10KRi5OJTVrL3BAISUkYjssbzkwaTUp SSZQImxPODtZR1tlNiRaZFZLIjd0UzNwYzx1dThKYHMpIVZjJytEZDgoMjAKTUlnJi9s aGFpXEkqQ2tHZlpMT008KT9bN2lVcFdsamw/UDAnZC5VSEM1QzRPXGIkLHMwUiNUIU40 dS9gXCxGSVoKczdnKV4zM1dyZVB1K3FRTTFGYktRKSNrNzRYU245X3NyZ1YyS19XP1hq QVVNJVpTX146OSM6WWJuKlpkT1pIJCwKPyQvUjw2RT1XSnBRN1xcUkxdZStvVTVxKSM7 KjBORlIuIjozRy45XT5Sc11UTlBMZWdBKTJKZyNmUnU3NWNUM0UKLTZAVnQ0aFdoaCxp YSk1Z3M5QDhdK1VjczotOE4rZD47biliPjdtJDUmVDRfaVphIl4pXTkiZzlvdGZGLkw/ I0AKT2JDL3U+YT9aQlw9NyI4OCE1PE9PT1xfRyc6TSFfX1QqNjs1UmEqTC9YYG10aGk8 S2NiPVNLTkJuc2hyXHNZZCYKL0UjQ0wwRF1NZmFQJUhHWFQnWUs+XVNJcFcpYUthbUpz JU5fXEUvM2xaZjIxQE1FZi0vckdAWUMtTCtBa1hoYFEKOzhbXFlcJ2RAMF8qaygnbjtS Vls4VlRyVUY+ViMyYVlRX3M/Pj9bTCNWNF1MQl84NmlCX0ZFYzksYjBhM3EqWlUKTzNk XHE+YkYkLEMxIyIxTDwlM3NYVFg6MlchST9dJ2IuakJXcDVba0tbcyZKIlcpWG1gPCgw Tj5EXSRtYiU0cTYKZV83XV9DOlxoX0pUUzJESGJibmo1YUAoNCJjYSZHLTQ4R2E4P0Ep NmROS2w3LCNeUm9mYil0dEw1OilLb0dFMD0KWWdIXEhxPV4wTEc4UTRqL1M4Ujs2UVs5 RFxGWWQ1clltRWtEb1tZL25FSGojIm9xTW45ayJ+Pg1lbmRzdHJlYW0NZW5kb2JqDTE2 IDAgb2JqDTw8IA0vVHlwZSAvUGFnZSANL1BhcmVudCAzOSAwIFIgDS9SZXNvdXJjZXMg MTcgMCBSIA0vQ29udGVudHMgMTggMCBSIA0vTWVkaWFCb3ggWyAwIDAgNTk1IDg0MiBd IA0vQ3JvcEJveCBbIDAgMCA1OTUgODQyIF0gDS9Sb3RhdGUgMCANPj4gDWVuZG9iag0x NyAwIG9iag08PCANL1Byb2NTZXQgWyAvUERGIC9UZXh0IF0gDS9Gb250IDw8IC9GMSAz NSAwIFIgL0YyIDQ2IDAgUiAvRjMgNDkgMCBSIC9GNCAzNCAwIFIgPj4gDS9FeHRHU3Rh dGUgPDwgL0dTMSA2NCAwIFIgPj4gDT4+IA1lbmRvYmoNMTggMCBvYmoNPDwgL0xlbmd0 aCA2MjUyIC9GaWx0ZXIgWyAvQVNDSUk4NURlY29kZSAvRmxhdGVEZWNvZGUgXSA+PiAN c3RyZWFtDQo4O1YuYWdOKSloJmNEZTFfbW1paCRgKDlEQW9STDZVNTFeNVslYGBJLVNx XzFHQHNLLD1DP2dqNmlEVCwmMUEnPApjTiFZc3E0W0ZsInEqW0coLDw+QzpBOGsmNSgy UlNbaGdeIytzRnFtaU0kbCxyIlZPNU0yMz9xSGI5WCdfbzdRJgpoYjNMZF8+V2EyMEU4 UU8xQjcoIiRgUmYjJSFhPi48OG1hbChrYUEzQmRWVmgwaSViZkhnPlRNKGRxW3RBKmFa dQpcW0JjI0lDUWEuSE1fT0ZGKUF0R14yTkRGVzlmTGAvLEBwXiVvVFNkYS5qN0c/WihB Ky4iWC4uaDE0Wzw2RG5BMQpoUGhkQ2ZYb0RRcCV1NU09NGFZNUQoXlA0aiJzPEIhPy1W RW8tMjhnTUJAaClNc0Z0QCtrKUQlYEk1b2BaMjYuUQpfKDMhKTJjUktZYFFwSi1dbTsz czRTcUwyN0UrPkxiJShFVlBkb1VmRFNRb002IUtnb11UaVplRVV0NWgyZko0QQpIWzE9 a3JgNGIvLihqPzBhaD9yYC0oM1BnLFkvVWRoK0Q+MS03UCosVjApO0lmNk9qOW8zbDRP REA/UCxtaydMSgozVyU/PWtmXiFtRkMtcjZTWF8iVElSO25QXCkuViVVIUcyWlZAOmdk LkJhTi45TFpWU2dLLVZQXVJLSFVlOidsJAphL2VFWV9rSEVXZ0EsT28lIUovNSpCTkZb SS0zJypiaUdgWic7b1k5VEElVl5jcWRCaE1TQFZgR0EsXmouIUNcYQohMCxyJ0hsLEsn Ul1cVS9QRHUuQGRNLWBIZEErKy1IcUpRa2Y8MDhoWXFyTGksRidvPDVgWHRtbSlROTE5 dGRtKwpQVlM/YiFeMC05KDxlKWxsSFpIO2ZLSk5XK0dydFZoVkBxREU0RTYiXDVkP3Qp KnNRPkhqSWlSWTk4WnUzXSUqWgosYExIPCE0Xm4tTyloczZyMV8uLCVzU3VlTCNLalZJ QDNBPUIkVC5lQz49SEQyNkBJOzFIXmteYV9aSSNTVnAuJgpLODY+QiElRCJjKjI4I09X LShoMjJDXkorODgxOlE8KypiNStaVy1pTS42MHNRQ2sqOGondD1BUmNDQzM4ZCZycQpt XFFybi40IzstWT4haUNZPFhQXSosQVklRlxPXylyZj1ZSDRkbyNgIk8nS2o5YTc8MF04 aiVnNylTNSw5KChbUgpZXVUiWjBRZG42ZmBNTkloVzZmPk9pOkU6VnEsVj8lLGAwbC5x MzJOIStmKCVCT006MFFwKSFPcF1ZMmw9aWdbPQppQz9FLTVuJ3VRVkFDTyhZY1w1UFRL KS5iXzhjMzFpQHFfKFY4LSErOE1nVEFaNTw2XUpGUVNKT04xc0hwRW1cQwpdc3JzMVBE I0lIWFtKdHJtMWBOW15OZG9ZVkBvZ1hZW2NhNGohXVZOZ0Q0YSpJS25kYXBTN09aQCw/ WCVJUiUhQQouIyonWE9hUihPQ0FQdVUwMF1tNyclLXJeNDZiPlBKbnFMQTI1XyhFPTZz QTNTaFZ1WDM9JzdjOm1ucFIubT0/SQopZGQnWl5dZlAyXE9BInAnL05Dbi0zYUhqM09L PClANmBCXzg9KTM6LUpscF9tbDw2T1VdQGp0MWlOZDplRUM6MQoxJWMoPk46OlgkZE9E TUQzaC0kJz90Lkp0J1pBMDU4bihqaGRTXmchUmkkXzxTb2RFc3JUJFVmKUAubF8rSnVz ZQpuIkxEaWwmOWdqRiJCUGJqcT5wSiZWZlRxVXAvOkxQNis1YFYtZF8lTzFPMG1bLkxP LWw7aDpWSFh0RWMrVltoSgpHQ2F1SHBxakpCSj8iRkpOKzk1M1NjXG09UFspV1Ulc25y PjttR2dzVkEhR1UyWWMoSU41MFw5WWxFRjshPCFxaApHLS9zPTEoNWZvZ0RXIjJoTzpn OUUjTnFzWHRIXSc1YysqIVE1QW8oQF5AOXJEZmcsSmVHI2lUVWVhKS5sPFhFLQpeWGg9 J0lxM1o9YHMsTlA6NG8qR1dVKFwiLXQrX2dGby1zI0lHLmVqPjQ8RmdiS1pnQjo6RyFe aDpBM0hXU3U1QwpcYiFPdFFmOHFmczNeO3BLW0dQQm5bRDBdZkFCLi5DYm0jUG0pX0Mk cF9HKFhaaUc7PEdLbktPOiMyNXQ0alohdAoiUnBkOFEpRTtvUFZyKyxCLzRCZGpNSz5F J2JmJEYuUV9YS2JWTFJJXk5JbSczRmNSPkRJT11UOystMFJlTTBSOgpgKCx1Uyg3WWI8 YD08JC1jLl9wWjVCZyctWilyUDAuZl1mLmRyamI4Zl4pKyVoUVspWi8tMmU1U28zbnVj QFRKQwplazg5aitwYG0+alJlW11jNVpwRFQoLktxJ1dCcicuNF8nYmYzZHA1bCtCa0I8 TkpkS0Z1KlVAPVReO3Mna1BnTgpXInJRZkVNXTBXZC1mVXBRa3RWRklyMnI4Q1lkJTY4 IVtFUjNra1lqOCU+RjhBTl8oYjIwVT8hPllfNE00YlZxdAo8KmsyOGRoUlYlNz86LT0+ UzQuY1EoTjBxcFo7ZSRdPlJoJzRCVDFkMmNUJCI+cHVWdDdvcE9sJzFgcyElOE8xaAon TUQxVWRPIyc2UzxUXzsvISUjRFE9MWosISJDKF9KJlsvOEBsMUUibk9pLEdFKDZzL3E0 bmN0PzgjODBuSkxBRwowdHBbRDAhZiFHMzEhaG9UIjcwcGQ6I1hBZE9xKzNPNjRsTmd1 cTNdVU8rZihlcmFOWiQwK1RjXUBNI2NJPl5HQgo3bCtRNDp0RXE5cTU8bC5HYyZlRmMj LEpJL1QqUFBTTU9XaWVWUlAhZ1wvOlU6K0VrOEtUYWFzP1M7c0FfbDw9Kwpnb1w9Kic5 JzYjUGJPciclUUdJYCMxPE5xTE5Ya1hFcXFoQic+ZFkqMFdXSjkyQkRkZUgnQU42PFE6 JlU1cDdzbgonNUtqT0BrPyYsOVkoNGlWRVBfaVpPMkVnLj4wZWBOaCdtOzMtJllhXz9P blY7OCooNWEpSW1WRi0ydUtBOW03IgpkQ2Ryb1tQLUFnJVdhPWBjQiVEI2E1aUFhYE9I NmdSXjRlZVk9ajNqTilrSWo7aGspYmomSmhQMEVDWyJwcUFBWQpuanVTcjovQF9qWEAj WEs2bkl1W209IzFgRUBnRGckKD5yZGVQM1ZxNWU1RCttXjFYRlZaZVg/T2lUPSghWlpt PApxaGUsckZDZ0o3JS8oa0MpXSMjXmkoc0tYbFdLPjMlLmJAYWBuPTxmOVNQKUpMP0Nm RWpCMy5ZczdwYUJqOkdyUwpoRkMydTMlc1NaUzpWTSxWQWBSYDIuSjkwYFBPQyFaJyk4 NSIlbCVwbnU+NkUuPiN1JyVtJ1RgZ0pacWAoYXQzVgo2YFo4PGgzWUA3SUhkSHJgR1Jd Z2RER1xlaF05MipgIkBRYWFAUTo6V24tJHVKIlklPW9OTE9LTjokS2ZNQFMhQgpFJyUv ZlwkKDcqSDJaa19wP3J1WTkoUDpWW2VDSFZKKVlHYFY3Mm4mYkFdc0dtWkZZRzZWdHMv PlNRZShNUF5rbwojbnA7PkMnVF9icjZCQT0oNDJYMi1tVWQwRHNiZj9uKHBfZHMrQ2ds KjItakJlMSNAMSxmaWszUy84TWBiXCI2bApXcEFrOl9Ra0RLU24pYylvbGEpdDVoMS9I XiVQPFxlMzQ4JzRSdEk+PkFCV0gqJEgvRFhtSjMhV3BKMilIS3NeLwptKDVvOkBWJEYm JTMtJGlXLU5wQi9mXzJpUGY1Z0M/UDk0cUAhTG5Jbk0/Vm49NC4mWzJpJW0yZGJDMWoi al1HOQowP0wnKVE7MTtJWEZmVks6RyJYPFBGZyUkITg4O0VlYDEpNjdPTSRkQEw1PEpK JWNOcGgxKU8hJl09SDBpI0ZmKQpDMm1wJTNmJlVabUNTXidoUzdqL1dkU2xgcGE/TU0o O18uJVRdJz5vNlV0YTotYEglW1ZDZltJUiFKY1MvOT5XawpmPmdmSSw0bWFwJiM8cDI6 RiFtMF43cjxVXDtJQzdgS2xiMFEmKFYqL0oxPF1tIXItRkVKLWc+LmNEanVuJ3RvSAo0 cHUqOFRVc29lQG9HNXFWLSsqJFpaSWo5KEBUaCFgNVkvPiE5ajIvMz46TUA4J0BCXWVd T09jY0Ywc2BVTXRrXApPLmFJOGlZKVRRVGVrYURVXVlwNGhuVjVfWCNMNipnLWJnLiEh MGFsXSpvI1NbNmNnQy5UXnFxRSgqKiFKMiRNMApcSUo5OEpNb1k5QUQ1LFNTZk03OmIr Nz1sK2Y+PD8pWCdfOmROK3ByUChGTXVfSUs2aCQ+QS0rP1VHZD1EJmNtOgpAdTk+Vl5u SDs0PitBU1AoSklWPyNvWlFaYz86cWAlKjdIQkBBSDtTOyhTbS8sZyZtbDU5OWRNR2gi K0xEJ1wnUgpwXlptYkAjSiJSMUVfNCc0PUgqSDtIQipsLzElby5laWszJSxmKXAkMS1m Tjopc2thR0JBZkoyVDcjSFghdDY7JgpbQj5bSjZaPDdZNGwsIlZiYXJAdGFIblA4Z1dR W18pLjMqZGouWT1IP2c/KScqZTZeZSYrLDVDNy1cKFQzcC9bQQonSmtNZTNtPz03ZC1y b09BRlRtVygwcztbZkdxZ15RQW9KNiRsSENUXD5BSmJDQT51QjBBVTRjU1QlYURgRShG cQoqdFJMVVksbDtBXHRUZFAqIjAzQyk9XD43O2s4Z2xnPTomSklWR24xJWprSzpyNV4o KDlQVygmO15OQDFHMlZnKQouKz9KPGRVRzUhUEgta1dpPmlIciEjWEc/cyEpPzNeLzNr Jih0NDBoOS0kJCxyRFNkXSMkQFpXXkEpLVRxQGw8YAo8QUlNR046YUYnWSxHZ0ZDRSwj ZW50OnBcKi9WdDchajFadVNJUk9NPT5tNSMiJV9WJzgnUnIyaj4/TTwvM3BYVQpCdFJ0 K1UhWT9sT2lVLTdYYjovXyVnVzFBI2YoOWVsVjgiWGFxVTEwO0A6QHQuYj1VU2RxZWdV Qy9gdUltP2VYSApVRS1tLVxSbTE+UDYlVHBeO0ViJzVCYFVJaU1yai5lW2JiOTRkWEUz ODNJPmNdWG47bEdNa0Y8OygoZ0EnUlNOVQpJR2NyU104KitxYzA4biE9KV5jO0sic1o7 XCxwQjFKVnQ/Zy9LcT4sVTknKV8kQT8oW1laOkZFS20wSk4icUhAPgo4QlBkTUouLmth ckZ1Ty9ZRj0uODpUXiRPR1kwWkFMW2RjdCNDWEskOWVbUXFecC5hV0crRmNjVURvUXRh Q1EoTwpjUiUiaGhqUyJNLUwqN2k+QUp0QG4qZ2FzbEA5Yjo2RTBscixsPkdIcFFEbSJr NTEjXWhoSi5xZS1IU2JtcVc0WwopYExyMnJGX1gnVUg7XGUxSiE/bkY0UFY1SU1Dc2k6 cHIxJidGRUguNTcpblBbMSc/Ymw1MSRsQUQvK1xRLUcuRApURTUoIzNEO2tKSHEkaWdj Q2Y8cmRVOVNKRmZSZiM3V102OSlnNDZdOiVrZ1JQI2E5dFJyTylJI2NDRW41cE9Ibgow OkwvJWVwLidKW2kuT2lbKm9FMF1cLjtWaTdkbnNWPUNeSU5yNCY7X0JHK29mZV8yZkNi cG9tbV8mOD1UP1MqRgpnPSIyVyRdO3FmbmVbaFBXbmosSVBiMEwwWSVdKkVdYVNmWkdv NjskLlNIbWQpIy9YYkYpMiMkKilERkYzX2F1NQpOZCZsSy1rNCVSL18kPU1LMUlWYSJV LCI1ZyhhMVlnb2pKR2xiN0orX1lhJDwrWTotMTIsPjplKTQwYkdQLDlzWQpZYTxwM3Fd OWVOPzVUTWVySyl1LFBCUz11LnFPZ2xOO1skIWxfLEJQOWlzODdBSXJJQnMvTzNrOCI5 bEhXaV1DUwpSVWA/cEk2Y0tlNGZGWmpENytqdG1fKzFBPGBqcTJFc14jdFcrKEIwKjw2 NVs/P1JpUjtBTyxOSnRzSERpZzFvbQo2RWQqXyVSLWc3Vl1kbG1mbSZEKj07XWtMRGM8 anJGakZYSGFZWSZHXHRHWyJCJD1XOE9OZmtqNS0hXGhULFRWVgprWEtFKS5edWxWZU5v XjhuLVA3QDEuSUZHL1kiWTEyQCpbbTheZExbPlhfU21wK01EZT1NKGQxPnVDKGYpT2Ax XAopSyt0OXBRW0E3YSxHXC4sW3BbR1IlYkJXKStpdGE/My5mVVheciJWZW8qWi1gSEku cD1IVG07Sio2XURSQEtWRgonODJVcDdTWlkyLkU0cjslMSNaa0w8TixPPW09anNIODpi ajUtMlAvYEloKiNzKEA7cGpxPnJcJFVWIkkldHA6Zwo2cUVYZDhpQ0BoaixsQUE/WCRI UTBOZWZJSEhzWF1oXkMnaTYrPTQ/Kl9hJSRmYF4lJC5KKFdGKEFoMGwva1hxPgo+WHNu Xl0hOjtTSHIubnA2WSVoJiNXNypjR1VZU0s1MCNRPz07aVdTIkE9PD8yJXIhMV1qV3FH X0tXbEQoOz5xWwowcFUraSZCZUpkQTotKWtMPCVMRFsxI0FyXjdVbipAYSo5XSpiXEJF cCplVWFPNiRMKFVFRl9qYDtNaTBgNSwmWAooW2IrbzJDXXIpZkEkR1JuN3NAXDk6LExH RXI+ZFdqXWcoW1dgPUAiWmo+SFtxanIhKD1wLWBUKzlGJi03XjkqOwpbS2FZW1Vwai9O ZVY0THRJRjdGNnBAS3NoZGRDNkQ3PnFtL2w1JiNmVjM2MkViOHNnR0RtbilwME1gTlFg WjJrRApTaSM/JCYrPm9Jaj5bVWBbZnElJFxua2FPW1w7XGU8SyNqLixmNjYkYXEuSmA5 VkpdQlZnNGU3LW5JPVBDWWMuIwpSaEZ1MDhBMW5KQ0xeSG43QVFwOzpjRyhsMW8nV2oi Q0ZbdDwoKT1CNVMhRSNjZkliKUhJPXJCWTolZHEpcGglMApbLURZP1Y7Q1AhUiMob2Nx NHIjanBeTiQoWjs+QVwqZj1SU0drS2xnVHUrdExHPChacFcqRyRyVWRFS18+TV5wXgo5 Ly41ST9DXWhYPltMZC0/QDRrOj0yNzk8TmskTHMyMTpfMVxeNCZiUkxwQismYT0kalNJ V1NYKmxIcS5qcCciSApQa2soUHBIU1dxbnUiUzw/ZnRUVnEwXj44RnBbXSJHK2ticSs1 TjRCVj0tb0lEdVtmYjdSKkxycHUrb10/ODJCVwpHZXJXdV0pV0hvYkhiQGpHVmE/WiFj SGxiMyF0bEMzUDMkJTNIVGtDVDNjbilzN2RFIypfNTFMW1dxaScoKig/YQo8TDNLOC1L cW9TTmt1a2kxcjZxNUdeYlFsXlc1NlguJj4vcmRSKCM3VyVfJGllPklkJENBZmJhUSRU bkIuU3RnRgpDPjQ0QkoxZkBWO2dkLmUtOVh0OFxYREdTM11yOypFOmYrT3JLZEUjbSJy OWIxN0c7J0w1LW45ZDZqQXNnXC1qMQpXJXVtaidcXEMxYS5rJW4uVCtoZzJAQj5fNUli bTJjdSdROGIlUyg5KUJSVFxtKS9oXGc1M1thSGg2Wlw+OUpHMwoiTWJcJiEiPzRYRVA9 XCdPQ2VPcForblhHUGYvQ3IhPEYrQTIiPkksPytVWTZgcGMyKmJSPV41cj1tRkohcGIy aAoxV1NScjAkJiJkQVlJaG1nPS5UazxyTlAkYkZDT3RDYXRANVtkNW9LPENrQlhLOlBE OjxQcDs6TlRZSzgwPD8nSgptdSMwQHFuIWBMYmtLXyRyX0c8XGMkWEN1NzlFKjtKMERf YUdZYExVY3A/ZkZbTEM+QSYzMSozZ2AvckZJLyhvKgo1QkdxMWRlZC4xKD1JUkw1S3E1 ZWlPLEU6ZTsqQUVNbFgxcEBAVjtUamJ1YUNbJltCM0tkKkMobmZTPVtgNGc/JwpfXig0 NWI0MSdMQzpCNDZsXTdBU25HXm5JcVZwWzNTbWdQV29VZmwyOmsnRk1NYSIsL2N0YzJT L1lcJTA7L0sndQo5UVhcVylSYlw/OGVsK15sU11HajJhVlMxNCM+SSRIXG5YbTVOdFEy ak9OJyVeKENVPCloaFJcUT5caWVhYS1hOgpPPE9XcENEOSxIN144ZllHY0dWJTYkM3NO Uys8JlcvJE9VaT5GTz8yWFE9S2JOIlVWcSFCVkRGNWZdcV07aF5FNApvczFbXkdPKCtY WzsiVTMpJXVLTCJXRmAiblhhb0A+PHNTUkU7am5ZVklMOWx+Pg1lbmRzdHJlYW0NZW5k b2JqDTE5IDAgb2JqDTw8IA0vVHlwZSAvUGFnZSANL1BhcmVudCAzOSAwIFIgDS9SZXNv dXJjZXMgMjAgMCBSIA0vQ29udGVudHMgMjEgMCBSIA0vTWVkaWFCb3ggWyAwIDAgNTk1 IDg0MiBdIA0vQ3JvcEJveCBbIDAgMCA1OTUgODQyIF0gDS9Sb3RhdGUgMCANPj4gDWVu ZG9iag0yMCAwIG9iag08PCANL1Byb2NTZXQgWyAvUERGIC9UZXh0IF0gDS9Gb250IDw8 IC9GMSAzNSAwIFIgL0YzIDQ5IDAgUiAvRjUgMzYgMCBSID4+IA0vRXh0R1N0YXRlIDw8 IC9HUzEgNjQgMCBSID4+IA0+PiANZW5kb2JqDTIxIDAgb2JqDTw8IC9MZW5ndGggNDg4 NCAvRmlsdGVyIFsgL0FTQ0lJODVEZWNvZGUgL0ZsYXRlRGVjb2RlIF0gPj4gDXN0cmVh bQ0KODtXUjQ/I25bWiZqQ204Sjs6KjpbM2EzJ15nRCRBNipaXTYxVWBnPUFeO111SCFo bz9AN21FKCwqYmA7T0kmdTAKQDRzPClnKUZuKG5PazNMQzdQXSZrVGBDT2xnQ3NnVD9n OC0/Wjg/UkNpdG8nITwtUF9BSipta2JUOU9dbm8vWEcKbnM1R28pWF81RzEmUSVOVj5d KSYnP11mYVwwUjphMGJfa0QiJFsoPiJKOkBaSjtNPHJvQitwKmMwSTxgcywtQDYKXV5N Jy0yPktwZFpFRlZAcXRkTkw4NyhFRGMsNzc8Z1diMTtlalc6S24mczZyPTRCRG9MRGA5 VVBDMHRvZidCJkYKYT50XklARU51Y1haIl8oLjV0OSxwWHEoTGhba2ZnK1pHUT1dQi8h YjNaWlAkIjBzSyEuZl03PywkdFYnSUBKaWEKcGlLWClScG1iclBlNVcyNDBndU1ucE5t LkRlPj8iRXV0cVpaTnVsayxQLUpKWC05Q2VUJSJdXG90RzciIU0yUkQKSCs+PF5EKmI9 Tk8yZFdeLzUlNlUlbjNsVmFSbktrcis+TVcxcU9gLkpRKFFyQl0rWyooJVptOWFGajpG QDQ/WEEKXk42P3NXaS1cUDQ4MW9TXTZDZ1woJj0qVmxFJWpjZCpmNktSXG1tYFFxcTVq KG4wNnRGK0RkRDNvbEU2XzxgSnIKQjFaJEUoPDliOUlkQyI6a2Y7TyRtbG0xY0ttKyFd YVs3RChmJU0sWi1eLE49T14mXV4wP2hxRFRxP1M/XUVnSVMKSVZkZnRxb10tXGY5QV4v L2lTL3QzPmwqMzhfa0ojaSdzUVVMdCxGck06S1NhWjpZbiVXXDBXbmg0W1lXTTRFRG0K Nm5iMWtEUzw9V08nN2hZQmY9bVJbO1pEbj9mN2glSmtJaklHIis/PydgJUA5bGQ1JT5l WixcVEJUcE01LDUzWG4KVk02Tzc+XiRrRSY6RU1sOEI4PTExQnQvdERFUk1uKCxOUURC MjR0JT9COyMsJiFESUBjPCRkZjMuITNrXSVxNGgKcW0rNFwnSFttUGBKWXVpbmZ1Y0BA Zik6MW5iQiZyWF5jT1NPMmB1J29fbXQ/VkttWiliKCJ0JSdVOHJgV2E5T2MKTypUN3VU ZSlLPkNMKiktT241UUIhMCYyZjdoP0dvcFxCO2xqcU1xP210MWBLaD4oOmw0IUAzdDNy Q147O0VrNE0KXFQpUzgxUyVnM1YrIV9ENik0NSIlKTQ2Pm9IOUxZLCNbYkAhQWQ2QE1T Qy4lMzw/VD5qRnUoZ3BiP1tvJltbLy0KZlxvPylrVDRVL29tWjJOcXFqZkRVT0M5UU49 JnNYLF9rSVI6a0I0WV9bXyoiM2E6KXApQE1nKUJGcjUsSSVTT2IKVyFiVUpNIihqTURe TVgnSz8vcFtDOTlaQj5BdEpRVGJdamFWRVlfQklZdVoqTFNbTj4rVV1ELiNPM2UsI244 PlYKVnRzJWNgQVhhUVdcOVVjZ1A/NzNHJmskUVpfSSs7PW5oTENtYSJAPExbMkYjVVFZ R25AXE4iJ1MpO0huRVgyOGQKWjFvZkMwdUwyZ1spZFo+L0ItLz88ZWRIVUBlZz0jQmRI KCU0LEFGK3FhMD9ZMFolOzY1KkRHMzhwTCoiRWAhSSoKU2ktR1RfUyJgYE43JF9KKVZo YT8sdS5mOG01Yjc8SDVwQyM4IV4haGdlRTZjaCMnZSskV09oT3FAJDhpQGg5WyIKIVNZ RjlqOXVbYyJYSSJBUl04O1srPWk3dUU6RTBxcG1NJVpeczJhIi9oazhTLz8/JypLdV1f cjskQE1lQFxOcjcKQGdRJkY6XjIjYzskbSNgNEkwKjhIWkNjJiFgSFhdYStecmEkNUxg Y1goJ0hyaS00YiVDI0Y/MnE6NUs1QF8pIjsKP0BFNFw/YnFhZSZGZFQhciNyTHQ5bil0 JitOa0VqOG9iUHMoYmxYblBXXDcuMF83VSxlaERnODYyWFheLS9zJikKIlhwVmUldFlR JEV1MVZtT0JPPkk7JSIoWlpQYnAyXTktMycycClhPlJKLWRHNFVwOD40SixjZ104amAw PmxdVkEKRk9Yc0xhU0JLUltHU1dIcCdnXjtPR2g7JzRjZFxLa3U1a20xSy4oaW9Ibkww YGZYaWQ4KzZTOW4vYmJRbU86TVMKKDRNRDBZMD5kUzJ0WiYqNG40ZmpoaW5wYjs7aTYh ZyRDcms4cCRKNE0wMjFOUV1OVT8/JTxQYTo8NDM0M15laDcKL0RPNSxcT3AtaSNkO0Bo Z1ksLlNRYDhxJGIvLSc/L0BFLC9uMGtLJiYvN1w9Tyg9OT1TYzpnYFM7Y1hmOlZpXlUK KWRqOXFRbGtGLkosTFpLX3MrMCMiTk9AWClpZUtNcEZgS2IxJyVzYDticDo/RVtTKDtl ZTpvYzRNNjw4KG02PTUKLltTRkhmM05FcTIjX1lDLk5rQXNLNjRmW2VSSEAkbll0ND8k K14nWmhoLVpfVT0vTkg1byk+ai0qNShUYDpRSE0KJjY3YEhNKDFDJVclc0hsLllFU1ZF Xjg9Wlc8SDokKSxNNDk+R1s5JlVRZ1ZCZStEb0RyLF5AZmIpIl91KlxeazYKMFclX2FM YD0nYGFsOF1MLCYvW2wyRTxTWyljNFxDVTRuMXNAYVRSV0MnWENHJ15cWi8hRCM4WW1C Rm1NLEY+W0IKaVonSmxfdGtNLGApYGBLL1wjM08+RlUkaz1DYDxfY1RIKWdnXDgiSydB XEJvW1U8MUJoOCtOQW06NWE6NC9lbisKV2JONU48SzYiKFdpRSYqSikiKClLOyY4KFtd MkEyW0Bqc1hnZkBGYykqTmdbV1hAKStOQyVwRW8mQTshPDxGOycKSU9JUT5wcDhCZUMy JWd1RkRzSTU6OU4ta2hASygwQywyN1BUK1YzbylMIyhqaU51XidbXFhQaT88UENHLlhK LyQKWTRvbW9ESURPKWZjQT4/XTonaUs8Vj8qN1Y5TVJAIl91RXMhZi5GVUBGLGNkTTxS ViQsNjg3W2AzU1EqY19uNlUKMGg2ayg1XzYmWkdSMWw+XVhdITxOYUxbOiMwNFVuST1m IWBNYlZIKiVOTllVbVJtMGpVLT1rUV4qP1dSVzFpXiUKK2c9M05PSjklOmtcVikjOEY6 RzJhcmkjPUR1OU9KZGMmcSgwIUdbJFZCI2dMKiFKNFhFVDJZMThgRWlFKlRHLyYKPms4 WSpfP2xIXixyMGhdTDlAPEw8KylsWjRSJF9sWCNbRyZxZiN1ZG1HU10zM2g0Llw3ImBN aihQZ2g2M2txKiwKL2drbCdfPjJAQTJeOTw3aio2VUUiY2E0RDBzSGc0LkI5Nm1JWGQ0 PGNUIzdqcF41JktFNGonWEhDdWdKRGJUWjAKKV1XNUZwXlxAQjM4V1hTQ1Y9X0NeWVJy Ri51c1lKbkk6THNqbU1CS09eTUlVK2k8OkI+W0JhQk00UTExX08pOnMKYFI4cSltPDEo PUFkaSNSVSg7Tlc/T3MhK2s1bm5FcVEiLD5LTltXR05xRXQ0JytLUSdNZ0l0TGVKLSFZ NVlCZkAKLUgrXkBCYTpGTDw4cWYyVVdHWS1TOmhoYzJJIV9fcSx0WTdlJ1kyRjduITVB U21QZ1o5Vi50PSNkZmhMVG8vYC8KWClATFpkXTZBaCU2NSxZbXU7WGgrKidVOWNFTGNe RkFuWl9fQj83RiMkSGpdM0wsUDw4Mkk4I0FDUTQ8IiZdTUEKN1ElLD04dEU/UGxwYWAr WlZkPSVSaDFGVU0jaVosMjokKFAmaik3RlxwciV0JTdlbjNHVydyZCs/WyVgTGhwYjwK Z0xAbzUzYFlvcCcqMzYkZURrZ2RnQW1oUiROWitLOGMjMktKMSJKSU1nXXEjcSE1P21V UmlzP08uQSFzXTU8a1kKciUwI2Y2Qyw0bSxkSSVEO2snWTAnIiY3OzNiZDt1aS4waERF QmxCTSM7aE5fNV9CQEpGP25PXFc8Sk9oKVhxRTIKQGxMNSYvKE4iP0o5Rjc/JTpeQClT MG0nSjJQKjwzYWtbXDQ+IzcnSmhZJTY7X2tDPk84Sk0zLiQvYmEnS01lW2oKSzUwNTg0 QFIjNzY4bVxERVt1b3VCTTk7dW1ZXSZ0QU0jcC1mJF1ZZ1FIJTlUKDhNdDQ+IzhFcktB J1VNMU00TC8KXFVcPls3YDBEJyw7RD1mU1B0M0QqYDVZbi5EZzdoQko7UDtmXjk9bE9C UEpRLDViPCFhO0hjWydtJyJdSD5zSkcKYS9fdDUjXUFYNCc/MmZoMU05Kj46NUEtMjZK Xy4vZ2s1ajU6Kk41P0VwQyNqYjxvUTk3QzxdOSJWZGNfTF0nU0gKNkMxM2ZnP3VncDFX R2kiYF0ubG9AcW1LTWszWmQpP3E8K20yZSo9Zy9aPU5nbTpLJTA3U2QtZSkxdVNoN1ZV UyIKKFtiZlsiITYoUHMwNWxeYV9nJydvRTtdLEIxOT9cWi5UZ1VFYjZVSmIpP3M7Oydf VmU8SDFXIj5JOS9jcUJSOVAKbyVZPzxibyhRTkZKQWpKISR1N249VjUyPiFPMisqcWBP cVddPGxcNFRXYz9RTVFbZGVBNjA5PUJLVU4iXS9uLXIKSGxjYmlZYUhHL2RnTlE1JTs9 Jl5DYiEyVSZVZylWVl9lLi5MW0o9UGY6OiVgcW9kQjYzRSpXYE04Mj8yKlhJNGMKNktG NSk0RzpkUTc0W1JLcyU9YmRIdFdSbz5uViVBXFk1RTY7KUwmaSlZZltRJFp1LjM4ZWAo KltIaCVqVTFjXjsKY0UjNFw1b2ZoQmxCJlNaUVdwQEM2R2xVOWUzT1dfcidWSjcrdC0u JDsmdTIsL0xlaDcnWUI2Mk03I25JVD1McFQKM09TRHQ9STQyIzk0QkxJQWFTO11kKEcp NFlWKUFccEpARWAtcFpFNVtfKDtNQFlbRTtJbyhcMjk8OiY1XSgzRzwKS2lIb08jOz4z b2B1RGdwVzlVWjgzXmgwSmpBZ2M7U3JhVTtNc0RUPFJVS3IhMVVdOlVnWEptbzFKU1dS L1BcXFcKX2ZMKisrMSNGbk9pL3N1bV9kUipPZHFXJ044Zz44S1RrN1Q6SkUoW01ucWU3 cElHNnVuRTVvNTh1R15uZnByZFoKXFMjUSVkVE4taGcxV3FwJlZIMkJwYEhcb28+Kk5H WmFZZmpDQyphWWtuO0VHU2MmLHJvIUNAXW06TF10bE8/STAKW1UmYFdCI2IoaUtfaShQ THVwcF1jNlg9QTtGYzZvb3RrVTVUWHVkXnI4aydGVlVZaTtKZGE3dStwMU1lKTIhInAK czNbOihAUicock00My5kOUwiSilxUGBcOEhdJyY8SzMhTmRwYlk3XE0xUW90JEVScC9Q XWlQaFU4M3MmOXNCIXIKW009RC1VKFNfNEBVMilOJUNYK0VjWWN0PDQ3OFM/PkY5UEFA Z0M5bW51aFtvXy8/Y21mbz1XcHJJPlcuWihdc2kKQ2InYkA0Y2xGMjdPYVpXbmg3RExe ckxmS2lnRGEobXBkbiRBMWlKcUpCWjpjbyoxITVrSlIsMHBfcy46VXJjWlkKcCJEdEZM PVYzTSU1N2okblZBLXQhTkIhW0BSRydJajM1RkplZFApa2JtPEUmOilfOiJOO05Lcikz K2VGKDZWLUQKVjBpYDltSSUtKS5LaS9EMXRFb2JGIkY2aDxxdSZUTnBSJmZbRyklPj9p TmMuQ0Y/UWkoTSxeMiNYdGRtYC5WUUEKRWckKWMoVzVMXG0tU0lLX0ZITiZhKWE5QyNW NUpMLC80KWsoOmcvQitATCJwSj0kXko8YElALytoLGwhJ00lal0KJChINjw3ZjVlZUsv MVNONW9JKCU7RllMSmIlTVguMGddQjdQWWlBMUVjXkc2JUUyY1JFWDQ2dGhFP21HaDol K0QKI0VrQGUtODVmbi1yY2gvZXRSVnVAWj9bVSdcYzBwODxxWWhKQV5WNm8tS2tmRkYu K1o8UT5vJlNFSW1MJkthQFMKayk4ZVpMIlwxNT0yaWJiRFJcOVskKFEkI1MiIzhMUD8j V2IkPDVyQCdlWFVhaUhrMFYzRktEO1UzVk1zTXRNdTYKK1c/PFgldEdtWVs7aUhVRG4m bnBhTG1nTTc7TzNoQzVYSEZEJjMwMWopK00hIjBKTWFJMVNccyphQDxLVSRUUWwKRyRs ZF1FME1CMWFwb0QmVmk7JXRPZUZSamUqcS8qKiZPbmEsMjpVP24yJVddSktbIj9dVVF0 WiQ9KTZRQlpsRDEKWy1kWm5DcURsOChCUW8kYD07TVRbaGRLaTtlTTFuVSs/UWU8JChV JnM0Jzd1bE5oZy1oaUJzLl4yIWEhQ2ZwLWkKbXAjRi0vajxxU0EnWmUnQ20kSG9MNyNj TSE9PTJnTGVuOV4jXD4vQTFhXUckVE02JS1hdSJKOEBmU2RHOyFlfj4NZW5kc3RyZWFt DWVuZG9iag0yMiAwIG9iag08PCANL1R5cGUgL1BhZ2UgDS9QYXJlbnQgMzkgMCBSIA0v UmVzb3VyY2VzIDIzIDAgUiANL0NvbnRlbnRzIDI0IDAgUiANL01lZGlhQm94IFsgMCAw IDU5NSA4NDIgXSANL0Nyb3BCb3ggWyAwIDAgNTk1IDg0MiBdIA0vUm90YXRlIDAgDT4+ IA1lbmRvYmoNMjMgMCBvYmoNPDwgDS9Qcm9jU2V0IFsgL1BERiAvVGV4dCBdIA0vRm9u dCA8PCAvRjEgMzUgMCBSIC9GMyA0OSAwIFIgL0Y0IDM0IDAgUiAvRjUgMzYgMCBSIC9G NiAzNyAwIFIgPj4gDS9FeHRHU3RhdGUgPDwgL0dTMSA2NCAwIFIgPj4gDT4+IA1lbmRv YmoNMjQgMCBvYmoNPDwgL0xlbmd0aCAzOTQ5IC9GaWx0ZXIgWyAvQVNDSUk4NURlY29k ZSAvRmxhdGVEZWNvZGUgXSA+PiANc3RyZWFtDQo4O1lodGwjJUheJlVqZXE2aVU4PSw9 PElGaCkhM1YhRDAwJEZKcEUwVitrPkwhXyIiaDFqSSo7XDUobWdKKmZHOwo1I2dBZDI6 TTtdNl8rSyYpQV02XjMqZmZxYTduLSNXOmciQytiRTAqPjwwUmkwMVJpOExiVWBUW1A3 Ui1ITXFfUApkJi8wUE8wNllqP1gkKzkxXHFOYCQybiVfbmhlT3EqK144LEQnZCpLXzVQ dTk1OV1rTi8vPEc6REk7WkBnLWIpawpCdHBqJEBaMXQqPWJlWWIiXUNMX3AvP3AsV1hQ ZnA1WUxEYlAvNlVaLTNedDBjdVFrMSldLmAncDlwQGQ8aFMzSgoiVGgtaV0vSkBkLl0o OERjSGk5W1htUUlUcj04aEFMXSdGXlM1UEJdUUExRklMUyQ8NEJiPDBrL0xhVWJPXHU8 VgptQmIjOC0uKWlgPU5CPzBXRF9qRDdLZiJJLllfcj0iJCc+RFJaX29sVEhbOiJaOSsh W3IrV01KcWgnYyJYNSYlXQpHZ0BqcFhiMmFzTGxGOVo6bmtQZSVVOl5uby5KME1yLiRp J18+X1BxXlUjTGkhbygxPVhbODtDL1cvNitVKnNOVgouJGYmXHFBITtcWHNYMFQoKDU0 SVQyI2RkYkxOQ2ovXlI1aF1UPCVrREBTRCxePyItKFklIy1xU0pBNjNmZTtiZwpgYiI3 NW5mIzFfKldFUys0Nl4mdUdrXWErXTclNVJES0NqImM4aCVoPjc9VEJbdTlkYD1EVyIy TklCY1JwT3UyMgo5XmojS1A4ODlyPWQ7XS5YL2hFKi5uTkgsXVtBLWo8UCYoTUskMUt1 MHBmcT9vMkllSTskT2NZUkljdXVkZT03WQpnWmBbPmsiMD0lOTo5NzZrbC5XTWhHL3I+ PWQ6KWxaZzU2bkhvPExQYFUiJ0xHTmBZT11PITFWWVo7TDddNTN1bQpwbi8qZ1BwZmoo Qm89YFBtS0FHQVFpImFMSHAncyJTSVQ1UVhTUmheRCU+Q2dGUmtZREwmT0NcWDZtNU1h Jk0+OQo7c20xMTJdWDAhal9FITZLV2tqI25HXlJYVVJiMCI7aiM7ZFgnXmFAUTpQSW8k dSQoV2QqUnM/SHVdJjBxLjBKPQpyKy9tJGF0dENuYWZzNCVYLDxSaVMnNHVAX1lZIzE6 PEEhL0xrTFxVbGBdVnBrdXA8VG9bZmdCPiNhOEk8Ill1LwpPKkZLdWBVKC9hZSNpZC1g RV41cilXT0xrPS4tR21eIzwsX1ctPFBDQ286Lm4lWjNcKUFmPDk6LTNFXDBvVidbNQo9 MEwhOGlTalQ/Y29qRFVfVFI7SURlYmFtMWFwRWIkKW9fJ2NlJ0xrJWFSUStWbENCYUxf Qz81Yit1MmhSO0NEWAozbDEzXzdIMlNGKD9wOGRvTTtCJGxrYFB0aFpwXiw/bzRHM1Ul Lklmbm11Q3U+TUlVaT0/a0RgWVJHbF9OLV1HVgoxMis/aDw3KmBqKk0hKVxLYVduYixw NGhlX19jblBnQ15xXmM0U15QQUkjLS9nO3VoajdBJUVibitybDNvOTJEMgpTR2tvYWpf MkFxP0pSNC1kLi1BVGhUNnFCbGxRWTBwJUxxXF5VY0Fpaz87ZWlAO2hIL0RbdWByL1ws VUZWQyEyTQohaEVsUFZnRVcxIzleOVQ+WDtmZCtrJipnL0oocy8tWDpwL0MhMC5NZTAl QStFM09oW1BUPjVJRS47dHRKN0dKMApIXFpdcWYkQ2NOM1ppcyYuPUpzYSJwJztNTUha aTwnRkdSUUFfSj04ZyooXU09MTRxQ0ZkUEpyJHNdNTEuUDA3PwpKLEo4cnBzJ2wyRSMq WHNqYzlsMy5scWpNWlt1MltgTSJQV2pwWUwtYkFsNFtNWTRWLCk5bWNZVUJzNmdMVEtV MQppW2o8KEcjdSdAVUJ0Mk08NjxjYktcJjRKa1Q8MS1wKjFlJVpAckY8OjUoSzUxOlpb RGEjXGsjZDFGQDdMO1wpQAojKzYxXT4sQVFXPE5iLXJRN2ROQDhgLU9hKy1GOF1jWz5S LUFrUSolbyhLUCkwQihiMjlPaWktU0pBNjMwZGQsWQpJWms5bkNqZG9iK1B0aj9IJEZX amxDRkdoSzJCTEFgUVQ/J2o8UXAqVikqXiVlbytvWyM7YSVra1VdLDlSX2ZXPwo+Jzpc MzBrVSNbMz1KVSRmMnIhPidUbWcrSFdVbXI2Q2MtdD9edS9dXi4tMyVOWk49K0I3Sls+ PF00PEcjW2c/VQo/cWM1LS02YTk3TnUrSkAnPFdeNTEvV1paJk9NbiFKaGRxXiFJV0Fo SS9nbDRBbSRNODZwUGZuOmtAIjA8Ryt1ZApxR1dRdVg3OiFkWSw4cWAoYE1OIWVhZ25g ckBANVBKK01lc2xyJkA3SyNjLEFLUDlKM2MvbXApITZXNUNtNHNVUgotI0lJYjVkUGhj MVY1aVY/MDg6cVFNRDhiVEtvM0wkQ2BANy01ciYvSFglLlZgJkdbPD5dSW9vLGJWL2pf OypAQgpBN0hyRUFLYTxwNCRYOC1bKkIiUV5hOm5CSTRIbVEiSUxFS1FtNHJBJktaIVtK ZnVoSyZJYzpXL1IoKStEJzlhZgo/aD84aSRrWFY/Xmk/W3IrZ08oPTRVc1U7Y1FzRkFz IzVOUVlfTydlZix0OCYmRUIhalwhRjwhRVZCKzdBTTtAbQotdWZ1U1AxRWMlaClBUFg3 SU1IPF4jb3Q5U3A6SGshZkBNWCJVTzdUQ1VZdD0xRCNkRCVsW2IzX2lgTVE1cjkhPwpn ciYiUkpvVThQQ2FVJ3I4KyhVWjF0NFE6NnRkRFljXCguU1hzXyIjP0dkTyJATEkraygv RUJxSyRlb05fMixPRQoyU247UD4mYkg+RzkzS0Q7YDlzSU9qPjxqbGc9OF4jdEZCbyJH OVg2LyJGTmtNcTIkXmQiV2ddaWVjSEInMGgkOgopUSU1PidxLWg5QmhcRDlcb1JqMVg5 TFVHS0gnKDQtS0dwTlJXY24sbSlrVl5BSnIxOj5Cam8xMVdURnEjJS9oOgovZ3BnOlo3 Jm1AVGluYz0qXlFbKUcoZFZgcEo4R0FYTWIlKWgpIlZwU1AtR3Q5VVtkaiVIKDFrJl5W JEcqRXE0UQpDaVlzVy1DTl1dQmNTUF9AazdVaFZJRkBNK1Y2X0gvXyFcTDZaI2w0S2oq V0g6XjlrazVwImwpTFZVS1hFYVg7WQooPkEyL1VbMzNOYUZlWWJmZFEvTD9SM2U6NlN0 InNvVjZQME9GMVEmNz1BOkIvKS1kTGxHR3FKLWIoOmtoJmhFZgpccXRIOEk2QUEibVhx OyRHTCs/X0coLyptbFRKVCNiMGw/KDlaL29kRl1LJmFBIUIhVjgoanRbTV9ccWk7Pio7 KQo1ZlhBOVJqXE5tZjQxTkE0dWddNnBQXzRKa0VoMTZoWGExMS8tOWQmbk1IdF88YWFU RGdsTzdyVipcSk9OWnJNWgpSJUhoYlhXW2xWVkBmXzJAbHIxNyw2WWxyTktSMXM2Vj1B Q1Y1Lyg/TGJjOSNrOkAtbUMsVU1wLG8qRERlQmxvWwpuRmNwPF9ZOS9yU21tUC9rWW5S YSJcMi5ISVpHV0RmN1FGMz9RMEdYMFN1XFFYZzgoQ1IkJjwzYitkKjM7ajwiYgo5KUwq MzgvOydaJmsrOStdYUwpdDFpXk01aEptYTlvIzlTYF84VHUrRUY+ZCtgR1MqY0dATy4i KzMsNFlhImFeJgohY0FqcS1ubmZrVVdmLFJLI184PUdXKzRBU2FEJ1NQQ3BcX0VuNiNE WmdNNEdSPzBJZEFfPj0qMXA6PS8lVkAsZQpDb2U7PVxgJlI5ZEBgNCYiVjI5dDlUcEQv PGZlXzxtRUcraUtONDJpS1k9PG48Jl09RiU/JjlRQnEmYGBSKiM4YwpBYiI5TEZdUVJj NmZeKEFqQS1ia0JdYigwV28+MHNsYD9gdEwsXlslMmRSUClxLWZFLihvNFJiJEY0KUJL RzhjYQpudD1GSyNTbGxnVmZjI08pZVY6RTdNYjRkQTpcVkMkTzFGWUs+MWMwIj9uPTVX J1k6OFZmYyRySjFpYCRpMTVzNgpsXjVhVT49VD5kQTVtJU0rY11dVSNzcE5iRUZVI2lQ VlF1bSxvTjlhN19OW1tmc2hgSSVCN3FFKHFpKiJwRkBJWQo2UUYtLEo4OD5AQ0R0JklO U3JOXTA+J0suZ01kP0Q5UWUwPWssUEsySCVVIWBLQSpvTXFQNkw0MDZtXXUrLXB1PQpH czQiPVU1ViZ1OypAcHVrWUU2Qi03LSdgTFBZMlpDW2liRl5mJzouO1E/YmlhZ1M8bisn VlxpMyY2TFs/TCtcQwpqWlN0dUhsJlcsNDZEV2BMOk5hYSgzQSZJUSVmLWkyWmg6NT5a RHVXLGNROCVRYUMqSThXOGI4K05RVDhXbnVLdApYNFU+VVAvbCc8bFxzYmRvblpccGxp KypELm1cdFIkMzc8N1EhXSo+aFBDKStoIzJTOlJrbidhYDspWj1jVHQpOwpFdEdJWzY/ WEZVV09ULlcjRj1uUW1HIkViWCh1SyFKWE1PRS1nMDs0cGE7SmJKImdFSWYvYkZcIzVC UUQzb2grMQo3Y14yZEIyLlNjUFhnM0lBK11kU280PS03MEZVQS5YNWpDYEFzLyJBQUwz RWQmdSZpZGkiIk9sTlBqRD9UNj1EWgpNV0pCdWZVXltgSF1OTlU9OUE9JE9kamFqYGNN LEElXkEqSTdES0BDXzlLWW86XVBDIy1nUWU5YlhkRlBRbj5eMwpMVk1dWWZZZG8pJzRc R0swPF08V1ZvTF1vMzcxVjlZKFVvWTRjSSpcJCtLalYhQmQiWTRTTiQxcFFALCRPRkVv YwpsQyFXWEBdNWxvOE8oVERvaiMrVERUQyFTSlZUYU1DJGU7ayt1a2M3RSFNLmhtXjQ2 KixWVUVkUXV1NiwiMzNJTApYRmk1NVsmX0hYcyRFcSlKSzAyKiIlYDNeNGBNZl1NPGZk NSNGMFMyJG8rZm9cRGpnY0M1YnFpPypQVDRNZW5oOAo7LiolJllTJCNlZjsub2Y+YSkh aic7JygxWWA7OjpKR1MtWz9eJT1mTCE9KmldR0dCSUhdTElmMlQxRTwqZVNxSQpfc2hZ LmMyQTpkS2o4bWJSZCJJQD5eX207Mm4iIjI6KiRFZGFgMTE1O1BFK0UhJUdDQ24sfj4N ZW5kc3RyZWFtDWVuZG9iag0yNSAwIG9iag08PCANL1R5cGUgL1BhZ2UgDS9QYXJlbnQg MzkgMCBSIA0vUmVzb3VyY2VzIDI2IDAgUiANL0NvbnRlbnRzIDI3IDAgUiANL01lZGlh Qm94IFsgMCAwIDU5NSA4NDIgXSANL0Nyb3BCb3ggWyAwIDAgNTk1IDg0MiBdIA0vUm90 YXRlIDAgDT4+IA1lbmRvYmoNMjYgMCBvYmoNPDwgDS9Qcm9jU2V0IFsgL1BERiAvVGV4 dCBdIA0vRm9udCA8PCAvRjEgMzUgMCBSIC9GMyA0OSAwIFIgL0Y1IDM2IDAgUiAvRjYg MzcgMCBSID4+IA0vRXh0R1N0YXRlIDw8IC9HUzEgNjQgMCBSID4+IA0+PiANZW5kb2Jq DTI3IDAgb2JqDTw8IC9MZW5ndGggNTM5NCAvRmlsdGVyIFsgL0FTQ0lJODVEZWNvZGUg L0ZsYXRlRGVjb2RlIF0gPj4gDXN0cmVhbQ0KODtYRUxsIyw2SCZxMG5xNmlSRTcpUiky TFpQYS9RZlNRK3UrQl9sPm5da1QrKGF1aURYJERIOEBsIVdbcW9lUj4KSC4rQT1fPSIw Z3ErLDwtQT9qY2VgT2BkYG81V2ZnJFxlcCpzLzF1KCUwJj5eM0NBYiFuWTFUXj8vcWpX LEQsPkEKJUBlKCs2ZlQsMTNyW2koI1gwLFVMO1VxTERHY3V0LjVSY01Iaj1TIi9HNzE4 UTw0aEtJU0dxJ0NNdFMpRkwwXScKLWQrKWZORD1wQUlmTTQ/TGBuPUcpOCkkNCYxV3Mt Nj1tJGosZ0M0PjIhQTFkcU5tdT5DXDRsWVNNXFlSXiVTW1UKSTwlQ0ZHXmFlKz5iL2Ui Z1RyV2lTVFpYQyFgPChSKmVXRz9pVipBK0ltdFFDVVRdSl1lM2giTUw9J2pqVEdfTGMK ND0oXFEpKm5YUyYhb3RhM0tXX01WYi10P2UmTyUjI0xdcF0nYlJYQjk5RzgoMSg9YkBG Wz9zSEduKClCKi05VWgKKy4pTCFhZEY4ND9Kb1cqPzVdTFpwbjBEW1NwI0RebkE9O19u Ml9HL2lHUiFgTmlqJTlJP0lkWzM4TS1TSzRoRzAKUmNhc3NlOTcnOiYvckssQDhJPV8s JEE1U1BrOCNCcXU5RUxNIU1eWSE9KjZsbjw4bUFJVHUkQzhcVFYqIShSbnIKZSFzO0td UDZLJSIrUzUnbilldHNMZU1OY1w8MmtYbCldXnVZYSI8TSI8KTtYQzpJQkooUis0al1v RzdxUTFsLFoKWixhP100UWAjWE9lXUkuJ1VRYjo5QEZTbDxrRChBWFpwPDlJbTlSXTlL cF8hWiM1a1FoNFEoQlhhLHFHOUgsUmcKKyJhNCdxVSFITz46dHNObT5IaysuITBBalEv NmkoQSpaInUmV1hFYjNMcCYoUzB1Mi5KUVpmRWp1TDo1OCJjIzQKPlBhM3E/ajVhaE1r ak41TSI5JlhiMFovKyFOLWdlIT9ZJnFnRUtrOkk9I1EoZylnNkJqZE1LTDwyR0dhMWZy ZmIKMkhUSyQ7KTUxaVBAMSEyRGVsSlgxNVg1bWIwNHAyUkJOP2IhQmhEWXMuNUYjVyVS QmMxUDc6WzVyWSZobDQ6cSsKRy1zKjkkbjtxO0ApIkpFXTcuNWlRIUI4ZSdQKlYxbWtv ZURZOE5vVio0KihRQUImSDowaEcqQV5sOWdaVjpCbUUKREJHTkxtTyozKlZIcnQ+aSY/ OEZedHJHTi8oQD8sRlZHV2s+VWBRcjBUb1NYR1NRbERha1I5N0Y+Y2VAMUpPXzAKWjQu SDc9UmVTXVdhbCsmO1dQKy0rVkwqYnFLYGFcYlpAUSJrR3Q9Y14rLzYrQTFxO1glISdI Xkk6ZzIvSjgiTHIKZl40NlJjNj1QU25xMyUuLXFZZEw/aExzXnEndT4uK0YtJ1wobk1z c2JzM0MzLjpHWmxpOzRSNmpTYnFgXigsUTsKWWhZOUZQL1NZdGlBbEwkMjN0TGYxMTYh W25Vc15aTzJCWk5qZ11DLVohRS4qUCdLMFUxcmElQ0VzL1ZdQG5BZCgKJFshI0FRYEZb PC07M1ZZPWNzVmMlRjVWU2VFRnM8SGVeRlBwREt0UT0hWz0vLnJjIS9MP0FoSSIxZl0m XjtpQ2gKSzw8NClZVmVPOVBmdTtoOGlaZDo+UEljRks3QzghSFtidEllPWo3N01ZMSxe XmdOUkIwRkdjQyRyalhZKy5AKXAKQUBDOjtkLVZOYUZ1Jy1bY2IlR2c/MVdPZ2g3YEZC OiVWIiRwT0ZlbyUqNDFkI1hvZyFGPF49dGJldGBzL3QvMUMKWG1MM1M2dFJjUGRjcD4u ZGhWXS9ZNzBNYjYhYWVFL25kLnVQPkVZPGBoSiZ0TD9GQFI0Oy5xRj84KzBuITEtKFcK UVVIWCYsL1UwUVt0LGNwNE4ma15TMzFUJ2BQUlU3SEFhTEVNWDA2Z3AxUVorIUUrXlw7 TVlsNUtMXm89SGtrIyoKbUoqLSwxa3VbaDRaU2gsLzIoPF0pXlVjdDpESj5HXWM9UyZb VnAkXiFhN2kvLXBcMjVbJTkyVEtwTVI8QlNUOGEKYEsjRyY1cC1tTVlYc0IuVykyUj5p ZW49SGIpTmNeKDdZXlRKY1BwIiNmRWU1NFpRNENLPCxMY15UP2I+bFtUOSEKPUtpV1w2 alVnWlpQJWAwWDVVTmBOZU8qO1EsajFycEcnRjE0PGY/Iil0WkRtSVQyOV8tY2F0Nz02 KXI5SXIpJ0MKYlRASS9rJF9JailGcDIhWVsnJE1AbjAvXFVGSj9jMHE4QGA2YEl0ajsm KlJYJFNiK14wPkpGQUVvczdgYjszJUkKKG9qN25IND4udStrWmNIYzAwZSsoUEI0ZjtX Y2UqLk0vXy1OWmItcGwhcy5NXjI7SV1fcEBXRDlHKC5xSHEpclMKOEJwOkFXOW9ZZ1tY V1JLcD9tUmhfcW5iO2sxKU80U1IvW01pP0Q7ZyJWTydPWSE9TkUrVnMqKkQxVWxbLzUv aFsKOExwQyFMZVBjUjBcQy5LUG5fN1tSKDk2VChfMiksQTYqM1lRLzZjTmRva2hsYUFo ZT5TXitHNDdPQitQWl1IZ1gKUjJFPissOHAwMGolX1xhXyxBUlQjXXJWMGNwOV9lXTNV V1pRMCIpWGh0KSsvWUAqK0pMYitJcUpfMUs7TDJDNlEKZStPbjFRMVRwaEM6dXBCJiE1 QjpQPDBHdDApUmIhUnRvbjpoSSJvTl5WRzY/SF9pNy5nUlMpUU42QCsnPWc0NlgKYmU8 TDxCPCJlTlImWSc3KEpqa2Y0LFRzYkxYS2NIbXVpRmcjVSg7U1x0IyteIk51S1BPJ286 NSpSQ3J1W2cwQDYKNzJfZ0EqUiMpUltudEIxV1BKIUNfLyRKVWM7XCVCOkdoUz1NJEB1 MkJVbSVnVzhUYWdlYTAhYlU/Y2swSS8/OzYKUjlyb1JOPkdVNEVzL1VKPCZwJ28xQFgq NSVBXEk/ajxQY1ElKm9OPSpwaUJgUUdcaCZyN21KPFQ3VmMoXCc0YisKSStbZkoibSFL UjFOLSVnSSg3akhCPC5sJGFqYm0wOVo6YmtgIyNHNWs7JCVQaE9GOV1RXydCZzhVYydI ajE4S1oKO21gK3BXTS1FSkguVUdyN1hIc1FsUHRsSWMpUGErQTlrWDclVTw5O0VZLmUk blJEbS9KQCs4cV9MbHFNKSE0W3MKRzhRKjZqPkNcJG0jSFE1MnBsYjcyL3Q/NS5nVyMv cm5GbWNfMT07Ilo5QS1hV2QwSWlxNE01aWFPYkZfIlhxWGQKTVRJIzZRY3FVVlByQ11a WVNUMV9pXWc+X25qJGNQN2srdG8mV1VTXzxTM3Etbk5eTi1OZlorIyJkP20mXiUhXkAK OW50RkJeSyZLY2RVUCdIMSFVWi4sPUUhZDBuSSM9QVJcOCFbNylBOGUtWl5kTjVuYitH YExeZjdQLURYLjcsUmIKPGIwQEUxQEVQRDxIXy1URXMvTzAoSWpcZTdcI2ZqYkwmTThh K0lBaGNRb3VQPExDQUZHYTFRZCdEaUU4cXJtKTkKPENJSVNJMDcoXm1IVj4+ISJ0IWI1 SF1jc2tWUHEkZDY2alxAXGZzKC5zXEJQUFFIMF1vP2RyQF89NGZmMy9dJjgKazM2ImBS aVc5KC5dN3BtRS0uWTReIzAkbywsNmx1Wj5gZXBSPHVhblIxRz05TkhkOTwkc24tLjoo aF0zOG9VUi0Ka15HJ0Q7VW46OGJeZUYyTSM6cS5HbEM2VWFxbSotRkYmU1FrZkAucV5x Imk9RkpwOTMobD85P19mR3NbKShRUVQKQk8tWF5SKDxqTi1cUjlVW1QsV3M+clZVQlI+ TlE9KnE/SldTOzU3bDNZRVdvSTkxXDtISmtxQkBSb3NnUyE9OTgKOV4vLjxdbFYiQy82 LC02YjtAOkRFWG5tWCd0dC0mL2UzPm4uVGZQQTgyKlcjOUtWQW1mOUxIImh1W3MmPEZ0 XDMKOSRFYGU4SDxoRjs+IV9AODRCZ0AoUlNHQ1BQI2BaJlczNDtRaCRARkhRYXI+PF5L OWBLT2lyMC8xOiYoMlVnSiMKaDYwbmhdQGczcS9SWjVVPWBAOWxQRFUkZWJwRCxwREVI W0kzVyhlOVxLOTxbQk1tIU5YODIkdSJyOGJlYkFBRScKN0lqQ25kVFdGXC1gSVtfWlAj M2hvI0gmM2dJckVVayomUmBvK14iVmVuQk0hOF1tM3FcS0orc0RFYHU/VFVEKE4KbyVe aDxvNW9iKjpLZy9gMXRnUDQ+LT0wbmdQPjA8XXBMQmFSZnRkTVYrQCUmJF5hImdwTicl clhYOTVyVkRoUDYKIlQnI0Q8N3E4NSUpLDJIPiMyc0ZDR10xdEI3SStfQlhZUFMmJDQ9 bUJRQFJpKFA1Z0kxUW1cJVppZEJYN1wtS3MKJDJAKDxYNV05aT49MipvWzU2PWQ7IUBX UUE2NnJsZ2RWa1FLKUVPJShOQklpX2AsK3I2V09hJ1o9VlpgL29fQXIKPyJ1Zlwrcjs0 c0U7bVgjYGVYSXFyWzY3UDFaV0x0ZmJUPldnIWE1NDxISV4sbz9BcG1BLjJQKCJNX0Vl W2dUdCoKOFwjZkw/YiVNc2lyNWdCKDRUdDouKmRBImI3QWpsX1VLPlwraUBgdF4kaXNS Q3RLbCdidT9jNDNCL21rV0laPz8KJGQlc0JvMUc+PjVWUjQpS28kL0xGWThXI05LTTUx NkY2SC4oL1NWRD9JdTxZMDdjL3FgcXVyRl0tJDtDIS1eMl8KNm9Fb2Q6ZDR0QCVpUkRj KlMwTSg5Y20tWE1uVjdHaDA+QUpcTT0vNz5YXWNoaSJPLWpwZ0M5aWc/VkNZTlE3YU4K YTtoTV5BXStQLF0tbGwqcmwwVHBeaEpjKVBtJD1UWEBnMkZUZFdMIUchJmspSURzK19T bSpaVTdBQShnSiJPPTIKRT5pU2BBaXR0aFFlMFZQNCQvTnFSVF1eOXJXUDRXWVhxKjlh Rz5JWjwxS1hCMCFaZHNWXVNfKk9hKChXcl9tNjsKT1clbDMqQT5QP0tsLkZvP0VgV2lx Ykk1dSVhcDk3az05J1NaQy1dXmRBYnFmWyouT0w/V1xFI2dKJkJVV3BfPygKQkY+J1RH Wz1DPiwiN1ducXVXUTtvRiZhWVQlYnQvMUt0ZEVnYyw7OmNeUT1rNkstOWk4OyRmaVU4 NDRhLzNYISUKOExFTDk9RWlyWCksb0VPO1NlcFUsWmhRZmgnO11eTWxqU0QmSCVCdS42 TEZfSzZqcU0uPm5PYVw+bjV0REdzYiIKP2tyL11uOChkU2xtK1xqVFJCNDpOXyJRN3Iq YWU9Um5WRC5ULl9uRkxUX0g9J1FcZl5ZL2hBOSNwJTM4SmtyTUgKOVQ0WDVtYFs6ZW83 ISFNaFxvP0tWYEolKVg0PU9vODtKRGZXXFQ0VD9fSzFmPD5VUmdkRG5rMVgvRE8rTDZY Zm8KIktzIy5gTjFUYzY2PEQtcSpCWic9Zjg5PDZPITwvVGBULXFKZzRCNS89aEFtWm0j UU0rS1dTMydsZWFYJTRsLFYKOmclNlhfLiYlWm5ManVSck8qYCtWKTFiWVQnIURhQklr MHBSPlA9LVwjQyVqVyM+cjBcMlsvKTxOSklaRSVpXkMKL1x1alRzLz5zNy8nIkFhQ1lO ITxvPyx1ZS9FVGktZWM3QWpCb2M8WVxbL2gnQiZUVjRhUmRwLCwpdT83X1pXV0sKPlE3 ST9KLXVtOUxNRWF0Uy1UWUNsOk1aXzdIY0xpK0tzYDNSP011ayhFYD1bSiNwKkRZLWsr ZkBkJ1whZV8hO1YKX2M+UWlZL1dPSUA5NnJfTkk0Pj4kc05qQXBnQyNVWWlaXyVkIkor XFdRR1swcTJVPio2MmorQVcmZiU/LGRJLSwKKkZCR3QkT2tpNjYuSFpCVTQqSExWPico PCwjLzQxUVJRb0VEI10zK2RxSkRjUTtHLVJyTzRjOkcxcCQ7a2RuN0cKT1wiZSRYPGwh JEVMVCllXW47XlxDb0ouLC83RiFpYT9KcS8sLjFnXkhiQ05aQzVhX2BCcWdvZG1eOTxr IjhvT0cKO0hRQ0U8Pkk2ZEVpXCVUYXQoOXE+YTF0Tl1XRW9ZTmFOJTFUWmVeQCpWbEcx RWdGLFdtaTZpSTpgRTVPZU91SEAKRydySSE6P1FSbkk2MWtZXz9WOWdDWlVKVS1nSVZp ZltPLjNZdGczN3A1VGA+bGk7K1RnMCctPCx1TylWX1N0VkYKTDdybiE4UDBhPyVaNGZs O0t0WlltYWZHT3IycCkrKTdDLWdZRCdtSm5IbS0pLUFkUWVvanE+cyQsJUpaLCRsYmwK QHJRLCgiYz5LSGNiYVRWazZBX2s1QWEkVC1pMHQ4OT5qV0BNTihHZTN1aDVLW0UiQ0hy SUVXWio7IlVXJjQ7PDwKaEtlTzMoTTVxTiJWWEVsTlhkRUkmZEUnalxPVEctQGhkOCw6 XGAwPF5PLWp1UjJmdTRWJSxqY00qYml1Jys7VE8KZj4jJUtxdS1laD1ILlBNNyVjK1Jm XztJMT9RPUJgWVEqZTxiYiI5LihwLWxzVksiIjNaXUImKFxBIT1pZ0slVWAKUSZBaWJZ QC0+N0trWGVWUCZnNXQyP1BKXEolQk05RVRFP2xcXzpPMz1eLHBuR2RINCZqX2xQLVQq XkY0LzkzZiEKa2s3UWk5RzJZbyRQQGBnUmRQOWRebUBTSiMsZDgyVXV0NSkqXV4jU24j Z1NZXjpxO1hBLV9ZW0kkRzF1amhnM0cKKkdKPFJqXjQ/a1BTS2UtRksxWFtSRWJHKzVA P1BHKU49QiZuSjw+ZGc7Lk1IPVFWK1k2I0NYJ0cuJjItKylrXSUKKlQnOi9GI25FNmBu QWs9YVFtRWBUcXNoVD81YlJDPV0yPlAhcV0waClAPGYmKVMpOl48MyZJTjksRGRBcSti LikKYCVQX25QNnByJSk6JS1bNSdIS1dnISFvPClGZFw9JixYSDNrMjw9YWw1c01pLDxV SzgoS0w/O3BYIVZlUVkmNHUKSmk5TFxmOTlyJlZMXGwjV2BfWD9hbztvXlwpYTc5ZyJo JlRsMkxiYSlaRkJwfj4NZW5kc3RyZWFtDWVuZG9iag0yOCAwIG9iag08PCANL1R5cGUg L1BhZ2UgDS9QYXJlbnQgNDEgMCBSIA0vUmVzb3VyY2VzIDI5IDAgUiANL0NvbnRlbnRz IDMwIDAgUiANL01lZGlhQm94IFsgMCAwIDU5NSA4NDIgXSANL0Nyb3BCb3ggWyAwIDAg NTk1IDg0MiBdIA0vUm90YXRlIDAgDT4+IA1lbmRvYmoNMjkgMCBvYmoNPDwgDS9Qcm9j U2V0IFsgL1BERiAvVGV4dCBdIA0vRm9udCA8PCAvRjIgNDYgMCBSIC9GMyA0OSAwIFIg L0Y0IDM0IDAgUiA+PiANL0V4dEdTdGF0ZSA8PCAvR1MxIDY0IDAgUiA+PiANPj4gDWVu ZG9iag0zMCAwIG9iag08PCAvTGVuZ3RoIDQ4NDUgL0ZpbHRlciBbIC9BU0NJSTg1RGVj b2RlIC9GbGF0ZURlY29kZSBdID4+IA1zdHJlYW0NCjg7Vl5xOTlTJk4mTnJHO1RnRzRR KlNoOytUO0FXXTRWOzBTMWBVRVovXlJtSic0VlRIOl1PInNKTiQnRUBYS1JHCl8iW0JC YlVPNjlHS19JaUJSSGcwNTstQlhyOkJkTWpoIldLQzQxUzNXVzwmaydwWiVdQSR0MCEi bVkoM2pua1o6Cm9qRnAoNUsmJWQ/L3QyR2YuL0tfPTZqNHQzME02aF9RLlt1Nl47OWIw REMpQUxqOT0rTURBTjRdJ25wLTEvNlhUCjRmWjE6YlsvbjJGM21ETVpjN1pcbGtjV0Y7 Z0RXcEleY1tZKTRiU2dlIVBRJVdidGVDMlkhT1Y5QGUlVlw1SS85Cl9ecUwiNUMmYWE9 Nm0/YURVYyRDclFzbSo9MDkzPEolLyszJGdFPHVNWDhLdU0ibT9kKU9qY2Vfa1c/ODo1 Omd0Cl9DVW08cC8+SlJUUkIkNldrRyJpVGMnYW9YM0dlOzxVRyIqVlpGRWItQWVAPl03 L0NOcW1MdWZeLyFEOy4ocjVPCjVQQFEpbDhEPzY0QjhrYVpSWjs6KEo2X2g+YWRdcmdm Sj4+Rlsmal1eVDMqTC1tREdcXitFZnRtXF4ubk5Ba3NLCnBtS2IsJz40YSooYWlAPl4s JENbLi5pazIjWVRNVlJIL0ROMysvZGk5cT49c0gvLyo0L1dsSWVeP14iKE0+QjdqCmlX PSYsbi9zNVZcTzVNJyEhJjtxXSVcPm9tM3RAb0dCSnM6SE42SkRyUzshUk5VViRMV0c0 bTBtc1dbTFNdMmJSCjJYODhhKyFpZGFrVTBmKVQrVmlaSyJNPGJGa3BTN3BfRztday5X JVcrcD0+dUhgYyVbaFZaPkAqdSVoaFsrdGJuCi11SlVhKTkpOUcpczlMP28iS1pQZUxM O2xEcE50Pyw7TyRQZkJhQCsxOiJ1OFMlOUxXOGpxTShHKTQwZyUkN0RbCi9zO2RCTzVr Smk6RTJAcFFBKlQwTXRPLWNXc3BnZ3BwbTxpJ3E9JSEuMCtpbD02cUFbJ0FuZkM6N2pa aW0raSZgClBOLWAhNEgiKiM5JzwsPWI3MktNPl5YRz9NYkAsSDYtLFs5SENNQkVJaFVt OWJRQydtTzFAdUMqM2Y7TDBxZixRCk1yWSU9a1ZBPl9MZCJdSiJuQDxpTyc4OUAldFI2 XGFWU09cNV9fZ10hWGBiVjg1TkpsWCxFJ2BvbVdsYDwqLi9jCl49PSE6UjlwcTgiVWV1 VyMvbyJJQEw2PkdubShtaVZPSGVLRENdNThcYiVZJ2BycCVjVVI3YjQ5WyZtPHFDREIp CkdjYjNqUyI0STBgZ0pvcydGTVJYVkdWTDdXITxXVi8xVktETnUnKHI/NSo4cEw7Pjwu U2UrXkQ+Vz4sPi9qJSR0CitHaSZAb0pqckU2NjBvPmwtRGtPIj4+SVhpVSpaKFREYkoy PyxyIU1WUSsnK1tHIzVtPlwqLXFRbkxmSj1SW1VLCkU9WydiaDVJcXRAL05VM19mSTsw SEVMLCM3VWhVPD9kQjlNUGMhJDptSXUjbkRXZC1KJTV1TnVYV0BFUzc4VlxWCmpwOHVp PztEdEpINzlwazpFb2tXaXJlJSpHOzdNOzYrPHNTUUZRMHBZVVpTVzAoTjwxRD5vO2Mr OVoiV0dTRGQxCkZPU2EoYE5FcVtyUFgtUElBdEooRmFyOnI0I25CaTMrIUFbXkh1LkhB aj1tQTIxOENDaUNXRjtKWFpjXT1KQC5DCmNKMj5GIlwrYTFlMi0iLDM3QiplKXVNOFxe Zy4tJzBaZiUpcCdlT19YbVYvKWgpUVVuY1VGUlxsajMvUSckVkNJCjorL1RZb1NvRiNG aXVUbyckZV1oX1NoRFhOciVDTmRyTTtxZTd0KDQ0V0c7JSlgWFZIOHRBI1E8cWYhcExi NUhdCjNyaklOZjZNYyZgZV9vcCE9aXNULCVkYE0rRC9DPmovPnRhUjliOU8mIldWbC9h Q2IsYWM9LllXZm9KOGNnW1JbCmw6XSUlUFdURSpePWMnR3IrdVFEV11gRGhmKlVYPU1g T1N0OGBNN3FuSiU5Z19XNjprU1xBUFBpMV8qRVdCTiZUCkp0bnM3TT40cUA5Ml9ASCEi Ty9nXUcuNidsSEB1IkRETWdKSk1BSSUjSjZiYVUtLzRmRlBXR0tvXE9rNyZfPF5OCi9S bDNaQUhvKzghTTdOOT0mQi9pQ1VwQ0lXcmEqOTkhPykyYSpyJXBfRk9FQEN0KUhqJixr c3EkYEEwOGA/VThQCjw0Nm8ob1xAZVFHRDleMUw2KitEcDlCMUMtKi03XTxGKjoxOE5e W0tAcl8uOl9gRmI5LGRIWys6Zm1jMGFBSlNaCjVxOFAjOS1sTzA/JDhzYDYkUjRgT2Qu YihcWnAvK1ZzNVIyUiE5MUM8O2hrPlpuWEUkJ1RfK1JPSTI0LUZAKmMxClMtcEsmXGZm OlNDNFxdTFpVViVsY0Emc0BjaWk8KS8hbUx1IyJib2ghO0kvPDAjbklcU1x0WC5aX1Bs NiREKS9sCjpvJGkwWk1WaEJHbk4uclF1MV9EVzApYyQlO1hYT2VjLF5jb0gqMD4hWnNW P140I3NIKnU0O1VxRHQ5MUV1b0RKCkBudWdcYiVfLFJPJjBKW2oqXikjQFctV0NCb2VO JFxePGdCUiNNZmo2XEphUm4wNnMoK0BWRFZfOFAkSkVoZy9ACkAnVWZUVylTTSI6VTVu clhNOkdEKHRRKCRSWWwnXWtkVCRHXy81bDc9SjVtWUVpcko4YWQ9PT1lcGshc1ItUVZc CiNmb2ViRzJQWGQsMyNiRlBDXSU2Tl8mJDchMSc9ZjdSLDhRWTNoR3Mia1dGaWZPJjsw UjIlIW5lPl10QjEwNWRiClJWWFxjZk5ROzVVV2khXUElOyg1SzdNP0U/RGEoXGBfbmdA KCRRZlonOkM8JUg7XllsWDQrLytWLE1AOUpxcGRoCmk5RyUtakF0az9MVipEQWRfTjlt Nl5zX2duXVdkRilLbXRuNVgwVmdbZCFZLjhgcSlRR01iYkVpNSlxKnJxIyRuCi5pTUdT SW1nWWs7OlQyTkEsQEhpMDlIYE9OMUZSdS1SLzAnQTddM1QuWW9AZWEvWShoamVDOkdM MCM4XWBsMHQlClJjVEdeRCFCV2IpQGs5Vyw6NS1QckpYRlZxVm1GWEctbWU3KClbc2dN SGtMN1ptOTlsUjNlUkpWMVM+alliaU5ZCkMwbTM1YzQ4ciY7PzMmPVdeMWU1TWtyT0FQ KUFsVEwmcDRYQFtGIU05M1JzPF80bzRLLC1nRU1vZkZcKzExOiZbCkUuSmhMZSxbLGJa Yi1oLUtwNG11YmE1ND1aSk1CcFtwLjVbPWxEOkFsLWVJRElxPzZWTjt1UjFmMS05cjAj SChJCitIJ2ByZD4xQjMiY0RONWtLQEM1QlNEMCxncy9rR0xATWImXylyR0U6I15INjlu WnIxZigiRWooWUxQQTBMRT9GCmAjbTohbFk8cHM0I1N0TTE7dVEuWDxiP00lZnM9RGk9 ZDIyPSpSIyFhUHVYTF0xQ25GSGAuPV4mMUNmUzphSURfCmc9YSxZS0NEKCFkWUdFbUg4 aSxiW09XUyopNiJmWSN1KXNvQUooUCpnJU1iZTVuXXE0R1VqdWoiKnRoYEZOXzwlCitW JUM5TXJRaUFJUjY1PjcjR1ZgYigmT19xSSRRLmIlOk9tYVwrO1YqLzA8LDxLTy03a2ci SygzN0hrbGs1YSwwCjFBQSNbIU5tRVciVy4qJlFfdSY8QlsxUlIsTDs3OzRHJlopKUZv WlZjYmpzUTFRU2pqP2BIZS9JW1U+M2JEMUBBCk0mTSgvX1guJWYkJlVgL2JgTC1EMEdj SkNFUEElTUwmbWFlPy8rdV46c1c+YG1sbChUaVVrJEJKP1UnamM/NytQClg3TWhjRWBn LDQ7KDFOWzQqVUhjKCxFPjM6Wjc6akRIKFlPZE1sMjZUaFUtRkAzTig9ZCFrclRiVTVy Njc0MWFxCmhKayUrbmV1QVJoVGtCaysqTkVOU3BYaW5GXzU3RS1aSnNpN0gxYUg7VG1A IzlkQUcqWkIuQ25jM2hObERvVzFHCk45QSlgVG91NTM7P3FAc2IpST43JGM8aDVWYGJm NVQhMlU4K3VVREFoc0RdUjFkTi9mOC9DZVZGMW0hNEZubGReCnJeUWthZHVXczllOWR1 IiFeTW90OTdfQDhNVmI4cCVnJi1QWW1mWlk5TmBKNiM7ZFt1XzhgW2c4OTQsRyQsL08p CmNQRz9VIlpoY0FCYmxXJUI5KlxQXW9pczUwY0dDWig2QmA+MzdjaFNVKU5cOFBRcWR0 ZU5XVj1wdEhmYFMqQy9PCjNsQVRGXiVKLFEray9AYCZgLkxFNjk1LitFWlhJY2ZgV0pM VEkydXVXOiwnO1xaYWw2LiVqJGJxMSQ4XkFbdVhzCk1MSWtqIU5UPDFNKmxlaC9FSjQl UGQwJTxpbFE7MidobEZaJzNgaUppQ1drVmpoSkEpVHE5Rjg6X2BUZEdcLFAwCk0wbVBe alVNJChcdWdlNm9zPzo+RWZdNFpUJT5KayJvRTdqTF1IJ0pAR2g4RWtqPC5sXicrKiVU P0lZajQuL2RcCjg6NFA/PGBoJy5PUmlcND1vUXNAV0huMV9NUzhpPT9uL04kLGpmLkNc cU9rZkloW2M4LEc2ViIxLyhmai9mMG87CjdmNCRYXClPUV4vZnNCUispIyE/TlpqUC4q TXFQZzsoIzxPQDtGMzVVYjwhT1pHK3UuMkFtOGpJXipfZSEjXHNmClZDWjxAVUZYY2FS Q0cqZ2EiNmhjQz5kRGdGbSF1M2lsQGUxZlpFZkglZjcyc2IlRDl1KDVMJGstQyRKITFV bm1zCjpZVHAnLjE2ZCNjPlNzSEovaiZTMmZQbSs8QDQsVFRFQ2E1SExnU2RuYWBQIVxS XWZxL1pULF9TTzEiWlQ2QV0/Cl0rJVhab3JtMHBmZjkrZz1hXVNxRFVZMi1IWDJEJz4k VnUmbCIzKm9CRmxpYjZVVCFhWjpINlxTY0k0X1EjK1E0CmlFJk1HakFZUkxxKTVbS2Mu UUZTYklUKW43ZEcyOEVCI005YikpKipQU0ZjVTxZVzwrYXRnMEE3LGEvM0VYZyJeCi9g OShYai8mXzdVajQuYGNPXz01UTxJSkpeMmloMF1pU2ZLKWptYjtJWGFUPS40bTQ1LitK T1pPdSdfcTRDMGhgCmdfIiNbbjtwYVwmRWRJbERETkBnWmRdbCVLZ01vcmcnQEdUbHFe cXRkbm05VD1VayZCcUUxcE9iZ09LLD4pJVJCClgsNHVlLykucFU7KSUmVSpVInA8YnRC cW8zciJNMWRNMVJyVlw9QkNfSVxRZVNLRV9eIitoY0NyRkklb1psQzksCi9pV0tsX0Jx dFlrczEqZUJqRFsqOUAiIy5mNTlLKCVPQ0dJMT0tcmBsQFUlJkhvUTorXyg9czpSNT0r LmNxYUt1Cmk6Q2E9VV1WbjVMUGktSF9pZk47YTFsW1xnRT02L01fbV04bTpQcCwxKjRR Mj0xIktoRSYlbyJnW1MtPTQ8JzRmCkVPPDItUGE0R1RdPl4sRyEiQ2FVLWhDa09eZ3Uw VWR1YmYkTkFGbjw0NiNJV1paQylPKmYjbWBFYXAvMWlQSnQoClt0ZiRBJy1fTlxvb1lv Qj48VVk7V0ppbWFuck1jczBZIjpLIkQ9WUoqNSQ7YGZUSkQsLiZUPCVgP184W2Vxa1de Cj0yV2dZP00nMG0kPHBtSFBRb2s0bkhwKiZiU0xoIlFhOE1zNCQrXGBxJi1nOzg7X000 OUYwTDQsOjkoSUlhbFlxCi1ycnBbRTpsNC1MbSI+LWFpKjpIUkReWCJbZ2xbYlR1XVBr LSxlZkdBPzs9VFlcITtIOmQ7OUszYmxpOUYuRFQiCmI+cGFCZ10nL1hbJTlPP2NgWSJd SlAxSVIuTz0oOk40R2lZLmBlRk5DZWo5aCtdNjpxQyRBIV9mOT10WkZfXyU+Cm5dbjhs IzJQLTYybFAmMWhvRkE/KyVzZz00Y2tqWWQhPTo2QDFVI2g2IiU/WUsjIVclSmtIKjwj RlpWXSRiPkItCl8nR2ZSVTk0b25hQzQlcG1PYS87MEtsXFFSKlxudSdOLEU3U15eQzRv OmhhTm04I29sQG9wdWZnW2s4OGImTV9kCj0wcmJOKktxREomNzkxQjYmZT9dW2RcTDIz Pj9bSyFHOUg5VT49YTYpZDYkcDpbRGlyJz5CNks7WU01cD0uVWkkCmwqXHE4REdqVSxN R3QlazBrLSlYXktfbV4kcV1bXyJAQitsZlAvLHRkKlxtJkZiOW5TWnJTYW1QLiUvIWU/ aThBCmI4aFFsTWtgTi9tSiQ/bnJZa2hxSj5iKH4+DWVuZHN0cmVhbQ1lbmRvYmoNMzEg MCBvYmoNPDwgDS9UeXBlIC9QYWdlIA0vUGFyZW50IDQxIDAgUiANL1Jlc291cmNlcyAz MiAwIFIgDS9Db250ZW50cyAzMyAwIFIgDS9NZWRpYUJveCBbIDAgMCA1OTUgODQyIF0g DS9Dcm9wQm94IFsgMCAwIDU5NSA4NDIgXSANL1JvdGF0ZSAwIA0+PiANZW5kb2JqDTMy IDAgb2JqDTw8IA0vUHJvY1NldCBbIC9QREYgL1RleHQgXSANL0ZvbnQgPDwgL0YyIDQ2 IDAgUiAvRjMgNDkgMCBSID4+IA0vRXh0R1N0YXRlIDw8IC9HUzEgNjQgMCBSID4+IA0+ PiANZW5kb2JqDTMzIDAgb2JqDTw8IC9MZW5ndGggMTUxMiAvRmlsdGVyIFsgL0FTQ0lJ ODVEZWNvZGUgL0ZsYXRlRGVjb2RlIF0gPj4gDXN0cmVhbQ0KODtZOU9iQXU+cSdTYDVG bXNwdC46PDNEPDA0SystQSUxYSFTUElEImRmaTAsKCxOVDwrQilrLTlmSCkvWGEtVF0K aS4vQzw5TyMsPVFkUS1eTCErXGNxL0hDYFNeKFF1RTVHL1dTZ0InLC1SIXJVJDBOUzlC OV8ydD9ASHQ9YU0iJD4KTzI0OTUqMWolQmJULGNrZStfNUJhM0g0aV1xTDYiTTFZdFRM WG80aFhYWlVEVGgoYUlqWDJKa0ctblMyKyQ5PEUKTCtmSTlVSiFibUJPXFFmaDVSKylr cTpiJDsldUNQb0pQIkJLa0FQQUdFMkUmZkEnWU4wNiNTQVZ0QFRoQ04zVTYKLHNwIys1 V1JgMl1sTl5aVm9VQl9mZ0ZnWG0pSFVqMlhOLGRYU1NqdTYyUkNpSUJGRzwyIigpODBT a183bFcvMXUKW1ZHOVNUKk9uPCosXWByaydJK2puZHNwXiteVkhbJUI+MWkzIzBkW2lW OTpEZSciaVo3NGp0YlpqTWhKMUkrJmcKK085I2ckdTJMRFZPZlpzWGxJPVtBV3NTdDFU MiwkU11fbUdta1c/KypLYVtja1dhSktyMT4wXClWa0JRTl5ATyoKOkpkMXNWTyktU2pG LF0+Ql9UPzJjVj1APkJePThLbVYlL1tVUWRNU05dVXFDLUVzLWBrZypBVF4hTSRMYUxa NUkKOWFyU0ZcQmxyUlUqbHNYYUlvaGM2MVFMKDhyTSQ+a0hkYicuRi5IYz5DPGZzZi5v XidFKyFIVGkxa000S0hUbiIKRDFYNCkwbmhzXUIkcS9IODViRmZyPCkwVkdyMS5MUi4x Xmk5SjFHLUpeMEBQIiozSUJLPC5gJytARSZWS2ZsTF0KJGdnME1RUXJWXiMuYio3VEhV Kk8lO0pMUTVYVkolV3NaLmdiVlVsbTUpPl8uKEUpP2hKMVV1MlE8RWcrPFtxREQKQUdb Ry9vNElmRUgmOW5hOzJUXStIJTxpXD1pclY0J0stdGElSzkkJjA9SWxlSkEtTyNMSSwr LTxWPD0sLi1HPiEKcikzMF1aIShiZSc8UGBFbDNMRDdfSitqW0h1RD5aJSJNXGFbLiFv QDdCUTtZO0FNQGFNPyYhJy5aP1c0T1VaYDkKJ2RIN3FwJExsM0xNSz1kSDBSV0ZsWipB W1htcGM5NV9WX2xtcyNhL00+KzR0IjU+cUJVN05VLmgkVk1EMGJTXi0KRkslOiROMyJl ZW5ocVQlW2JwRnRTUEZ1ayUrQW1gbF9SYS5Ya0ozU0QmJyJhJnEpRjgrYWFEUkNgSjhS LVY1MT8KK3E2LScvcmNNPydQYVVgOTMjXzwrRFw3KmdTKiosbVJzTSI2YjVzTzxSP3M1 SzAuXz9JajUrKmZGKyZwTTROOScKckhwOExcMEA9ZlRiSi1EJk02TyJFTChgS207c0Ri KHJQUUhnWy9PXkl0V0NfVDUoOjJNQjFaLm82Y2tqZmNjREYKMlAzRT1ZUkc4SlMkcSso I190bT5TNVJxJ2ZVPVd1TkNRPStnO18/az1xKVI9TkZQclMoUkEuODMiUkokQ2ZAJzsK QEJYbG1nSGlDRTE5OWBCUFU5c1QwajptMWlLJFc8cyUjOzVJKCU8MWU/SFlSXT9XYTRO WyhLLkRadHEjcilmSisKaF1TRF88Oz9Pa29yKzg9NU9XWjwtMFQsKklHLmZdYW4/J1xA Jz9kQW1ZIlExITU4dGdyK0dZKWM/WEotIjlpKF8KVF9XaCcxL2Y4XGhgNVc8TiwkKjc+ TG9bMXBAbSIkXFVAZVVbI10mIzZpWS5pbipSZj8na1Fzc1A/bnAyVkkoQGUKKyd0ZXA0 IjZVPG9XZnJKPFM7dVxdOF1SI01TKyJQNj5MLTVpbTBRaEQ4dDFkR0I3Z1ZUMERPJzMr UHFnbl8nXyIKPTprRFRxZF1ROm1nMUFzR1g+LUtIWVl0RzBIdSdbWFlsJE9AIyFCUC1V JklIM2g/NDAhKCUwZS9Ifj4NZW5kc3RyZWFtDWVuZG9iag0zNCAwIG9iag08PCANL1R5 cGUgL0ZvbnQgDS9TdWJ0eXBlIC9UeXBlMSANL0VuY29kaW5nIC9NYWNSb21hbkVuY29k aW5nIA0vQmFzZUZvbnQgL1RpbWVzLUl0YWxpYyANPj4gDWVuZG9iag0zNSAwIG9iag08 PCANL1R5cGUgL0ZvbnQgDS9TdWJ0eXBlIC9UeXBlMSANL0VuY29kaW5nIC9NYWNSb21h bkVuY29kaW5nIA0vQmFzZUZvbnQgL0NvdXJpZXIgDT4+IA1lbmRvYmoNMzYgMCBvYmoN PDwgDS9UeXBlIC9Gb250IA0vU3VidHlwZSAvVHlwZTEgDS9FbmNvZGluZyAvTWFjUm9t YW5FbmNvZGluZyANL0Jhc2VGb250IC9Db3VyaWVyLUJvbGQgDT4+IA1lbmRvYmoNMzcg MCBvYmoNPDwgDS9UeXBlIC9Gb250IA0vU3VidHlwZSAvVHlwZTEgDS9FbmNvZGluZyAv TWFjUm9tYW5FbmNvZGluZyANL0Jhc2VGb250IC9Db3VyaWVyLU9ibGlxdWUgDT4+IA1l bmRvYmoNMzggMCBvYmoNPDwgDS9DcmVhdGlvbkRhdGUgKEQ6MjAwMjAxMzAwOTM5MzEp DS9Qcm9kdWNlciAoQWNyb2JhdCBEaXN0aWxsZXIgNC4wNSBmb3IgTWFjaW50b3NoKQ0v Q3JlYXRvciAoTWljcm9zb2Z0IFdvcmQ6IExhc2VyV3JpdGVyIDggRlUxLTguNykNL0F1 dGhvciAoQ2hyaXN0aWFuLkJvaXRldCkNL1RpdGxlIChNaWNyb3NvZnQgV29yZCAtIFRB TE4tOTcuSlBHJkNCLmRvYykNL01vZERhdGUgKEQ6MjAwMjAxMzAwOTM5NDMrMDEnMDAn KQ0+PiANZW5kb2JqDTM5IDAgb2JqDTw8IA0vVHlwZSAvUGFnZXMgDS9LaWRzIFsgNDQg MCBSIDEgMCBSIDQgMCBSIDcgMCBSIDEwIDAgUiAxMyAwIFIgMTYgMCBSIDE5IDAgUiAy MiAwIFIgMjUgMCBSIA1dIA0vQ291bnQgMTAgDS9QYXJlbnQgNDAgMCBSIA0+PiANZW5k b2JqDTQwIDAgb2JqDTw8IA0vVHlwZSAvUGFnZXMgDS9LaWRzIFsgMzkgMCBSIDQxIDAg UiBdIA0vQ291bnQgMTIgDT4+IA1lbmRvYmoNNDEgMCBvYmoNPDwgDS9UeXBlIC9QYWdl cyANL0tpZHMgWyAyOCAwIFIgMzEgMCBSIF0gDS9Db3VudCAyIA0vUGFyZW50IDQwIDAg UiANPj4gDWVuZG9iag14cmVmDTAgNDIgDTAwMDAwMDAwMDAgNjU1MzUgZiANMDAwMDAw ODEyMCAwMDAwMCBuIA0wMDAwMDA4MjcxIDAwMDAwIG4gDTAwMDAwMDgzOTUgMDAwMDAg biANMDAwMDAxNTIxMCAwMDAwMCBuIA0wMDAwMDE1MzYxIDAwMDAwIG4gDTAwMDAwMTU0 OTYgMDAwMDAgbiANMDAwMDAyMTQ3NSAwMDAwMCBuIA0wMDAwMDIxNjI2IDAwMDAwIG4g DTAwMDAwMjE3NTAgMDAwMDAgbiANMDAwMDAyNzA0MiAwMDAwMCBuIA0wMDAwMDI3MTk2 IDAwMDAwIG4gDTAwMDAwMjczMzIgMDAwMDAgbiANMDAwMDAzMjQ2OCAwMDAwMCBuIA0w MDAwMDMyNjIyIDAwMDAwIG4gDTAwMDAwMzI3NTggMDAwMDAgbiANMDAwMDAzODQ1MiAw MDAwMCBuIA0wMDAwMDM4NjA2IDAwMDAwIG4gDTAwMDAwMzg3NDIgMDAwMDAgbiANMDAw MDA0NTA4OCAwMDAwMCBuIA0wMDAwMDQ1MjQyIDAwMDAwIG4gDTAwMDAwNDUzNjcgMDAw MDAgbiANMDAwMDA1MDM0NSAwMDAwMCBuIA0wMDAwMDUwNDk5IDAwMDAwIG4gDTAwMDAw NTA2NDYgMDAwMDAgbiANMDAwMDA1NDY4OSAwMDAwMCBuIA0wMDAwMDU0ODQzIDAwMDAw IG4gDTAwMDAwNTQ5NzkgMDAwMDAgbiANMDAwMDA2MDQ2NyAwMDAwMCBuIA0wMDAwMDYw NjIxIDAwMDAwIG4gDTAwMDAwNjA3NDYgMDAwMDAgbiANMDAwMDA2NTY4NSAwMDAwMCBu IA0wMDAwMDY1ODM5IDAwMDAwIG4gDTAwMDAwNjU5NTMgMDAwMDAgbiANMDAwMDA2NzU1 OSAwMDAwMCBuIA0wMDAwMDY3NjY3IDAwMDAwIG4gDTAwMDAwNjc3NzAgMDAwMDAgbiAN MDAwMDA2Nzg3OCAwMDAwMCBuIA0wMDAwMDY3OTg5IDAwMDAwIG4gDTAwMDAwNjgyNTEg MDAwMDAgbiANMDAwMDA2ODM5NSAwMDAwMCBuIA0wMDAwMDY4NDY5IDAwMDAwIG4gDXRy YWlsZXINPDwNL1NpemUgNDINL0lEWzw2OTc4NjgxNzE2NzZlZWZmYTY5MDNhOTIwZmI5 NjQzNj48Njk3ODY4MTcxNjc2ZWVmZmE2OTAzYTkyMGZiOTY0MzY+XQ0+Pg1zdGFydHhy ZWYNMTczDSUlRU9GDQ== --============_-1159029179==_============-- From WJHutchins@compuserve.com Mon May 19 11:56:35 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Mon, 19 May 2003 06:56:35 -0400 Subject: [MT-List] unknown words in MT-systems Message-ID: <200305190656_MC3-1-3969-FC03@compuserve.com> This is a MIME-encapsulated message --cde99add-5810-4a66-b34b-dcdf24d6fdea Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline Dear Bruno, A few more references for you attached John Hutchins 19 May --cde99add-5810-4a66-b34b-dcdf24d6fdea Content-Type: text/plain; charset=ISO-8859-1; name="Unknown-wds.txt" Content-Transfer-Encoding: quoted-printable Content-Disposition: attachment; filename="Unknown-wds.txt" Unknown words - references Luckhardt, Heinz Dirk. Automatische Segmentierung und =FCbersetzung unbek= annter russischer W=F6rter. In: Lexikon und Morphologie als Grundlage ein= er automatischen Satzanalyse. Saarbr=FCcken: Univ.des Saarlandes; 1975: 3= 4-50. = B=E9mova, Alla; Kubon, Vladislav Czech-to-Russian transducing dictionary.= In: Karlgren, H., ed. Coling-90: papers presented to the 13th Internatio= nal Conference on Computational Linguistics...; 1990 Aug; Helsinki, Finla= nd. Helsinki: Yliopistopaino; 1990; 3: 314-316. ISBN: 952-90-2028-7. Reeder, Flo; Loehr, Dan. Finding the right words: an analysis of not-tran= slated words in machine translation. In: Farwell, David; Gerber, Laurie; = Hovy, Eduard, eds. Machine translation and the information soup. Third Co= nference of the Association for Machine Translation in the Americas, AMTA= '98. Proceedings; 1998 Oct 28; Langhorne, PA, USA. Berlin: Springer; 1998= : 356-363. (Lecture Notes in Artificial Intelligence. v. 1529). ISBN: 3-5= 40-65259-0.=0D Gdaniec, Claudia; Manadise, Esm=82; McCord, Michael C. Derivational morph= ology to the rescue: how it can help resolve unfound words in MT. In: Mae= gaard, Bente, ed. MT Summit VIII: machine translation in the information = age. Proceedings; 2001 Sep 18; Santiago de Compostela, Galicia, Spain. [C= openhagen]: [Center for Sprogteknologi, for EAMT]; 2001: 129-131. --cde99add-5810-4a66-b34b-dcdf24d6fdea-- From Angelo.Torquati@cec.eu.int Mon May 19 15:05:43 2003 From: Angelo.Torquati@cec.eu.int (Angelo.Torquati@cec.eu.int) Date: Mon, 19 May 2003 16:05:43 +0200 Subject: [MT-List] EC SYSTRAN: THE COMMISSION'S MACHINE TRANSLATION SYSTEM Message-ID: This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01C31E0F.BCAFDAD0 Content-Type: text/plain; charset="iso-8859-1" Dear MT-List, Many MT-List subscribers have contacted me to receive the article about EC SYSTRAN: THE COMMISSION'S MACHINE TRANSLATION SYSTEM. This article is now to be found on Europa, the portal site of the European Union, at http://europa.eu.int/comm/translation/reading/articles/tools_and_workflow_en .htm#mt which includes a lot of other interesting language-related articles and speeches. Best regards. _____ Angelo TORQUATI EUROPEAN COMMISSION TRANSLATION SERVICE RL Directorate - Resources and language support Analysis of needs and multilingual tools JMO B4-70 L-2920 Luxembourg * +352 (0)4301 33960 Fax +352 (0)4301 34069 * angelo.torquati@cec.eu.int ------_=_NextPart_001_01C31E0F.BCAFDAD0 Content-Type: text/html; charset="iso-8859-1"
Dear MT-List,
 
Many MT-List subscribers have contacted me to receive the article about EC SYSTRAN: THE COMMISSION'S MACHINE TRANSLATION SYSTEM. This article is now to be found on Europa, the portal site of the European Union, at http://europa.eu.int/comm/translation/reading/articles/tools_and_workflow_en.htm#mt which includes a lot of other interesting language-related articles and speeches.
Best regards.
Angelo TORQUATI
 
EUROPEAN COMMISSION
TRANSLATION SERVICE
RL Directorate - Resources and language support
Analysis of needs and multilingual tools
 
JMO B4-70 L-2920 Luxembourg
( +352 (0)4301 33960
Fax +352 (0)4301 34069
+ angelo.torquati@cec.eu.int
 
------_=_NextPart_001_01C31E0F.BCAFDAD0-- From harold.somers@umist.ac.uk Mon May 19 10:25:03 2003 From: harold.somers@umist.ac.uk (Harold Somers) Date: Mon, 19 May 2003 10:25:03 +0100 Subject: [MT-List] Teaching Machine Translation Workshop: deadline not yet passed Message-ID: Workshop on Teaching Translation Technologies and Tools at MT Summit, New Orleans, Saturday 27th September 2003 This is a reminder and confirmation that the deadline for submission of papers to the forthcoming Workshop on Teaching Translation Technologies and Tools is June 1st (and not May 11th as stated on the MT Summit website). For more details about the Summit, see http://www.amtaweb.org/summit/ For more details about the workshop, see http://www.dlsi.ua.es/~mlf/t4/ We would especially like to encourage colleagues involved in teaching translation tools to students of translation studies to participate, even more especially if they are located in the Americas! ======================================== Harold Somers Professor of Language Engineering, UMIST From debe@comp.leeds.ac.uk Wed May 21 14:06:21 2003 From: debe@comp.leeds.ac.uk (D Elliott) Date: Wed, 21 May 2003 14:06:21 +0100 (BST) Subject: [MT-List] Parallel texts for machine translation evaluation Message-ID: Date: Wed, 21 May 2003 13:33:07 +0100 (BST) From: D Elliott Subject: Parallel texts for machine translation evaluation Dear all I am collecting parallel texts for a corpus designed specifically for MT evaluation (to be made available online for research) and would appreciate any advice on where to find parallel texts (or translation memory) of a particular kind..... Source texts/extracts of approx. 400 words each in: French, Italian, German, Spanish, Chinese (Simplified and/or Traditional), Japanese, Russian and Portuguese. The challenge is that these must have very good quality human English translations which can be used as a 'gold standard' against which we can compare MT output. (NB British English if possible) I am just beginning to realise how difficult a task I have set myself! (Another problem is that some multilingual sites are localised to such an extent that parts have been rewritten rather than translated - doh!) An excellent source for my needs would be translation memory if anyone is able to provide me with TM and copyright permission. The kinds of texts in the corpus will represent current MT use. The following (provisional) categories have been selected, following a worldwide survey of MT users: Technical documents (eg. software user manuals, online help, telecoms, automotive, aerospace) Correspondence (letter/emails) Academic papers Tourist/travel information Newspaper articles Medical documents Scientific documents Financial documents (stock exchange reports, banking, insurance) Legal documents (including patents) Calls for tender Internal company documents (eg. minutes, training material, company reports) Any URLs or other sources (even on paper!) would be gratefully received. Sources which do not require copyright permission would also be a big time-saver. All sources will obviously be acknowledged in the corpus. I will post a summary of feedback as soon as the deluge stops (wishful thinking!) Debbie Elliott For more information on the project so far, see: Elliott, Debbie; Hartley, Anthony; Atwell, Eric. Rationale for a multilingual corpus for machine translation evaluation in: Archer, D, Rayson, P, Wilson, A & McEnery, T (editors) Proceedings of CL2003: International Conference on Corpus Linguistics, pp. 191-200 Lancaster University. 2003. *************************************************** Debbie Elliott Computer Vision and Language Research Group, School of Computing, University of Leeds, Leeds LS2 9JT United Kingdom. Tel: 0113 3436818 Email: debe@comp.leeds.ac.uk *************************************************** From naa@mimos.my Thu May 29 05:53:16 2003 From: naa@mimos.my (Dr Normaziah Abdul Aziz) Date: Thu, 29 May 2003 12:53:16 +0800 Subject: [MT-List] selecting almost similar sentence in a parallel text database Message-ID: <3ED5923C.20EA243E@mimos.my> Dear MT list members, Recently, there were active discussions and exchange of ideas on Example-based Machine Translation (EBMT) issues in this mailing list. I found that was interesting and useful discussions. In relation to EBMT, I am looking into various methods of selecting the best or closest similar sentences from examples stored in the parallel text database. Has any of you worked on this and/or could give some pointers on such selection methodology that is acceptablely functioning for English language. Thanks, Dr. Normaziah MIMOS Malaysia From carl@iai.uni-sb.de Thu May 29 17:57:36 2003 From: carl@iai.uni-sb.de (Michael Carl) Date: Thu, 29 May 2003 18:57:36 +0200 Subject: [MT-List] selecting almost similar sentence in a parallel text database References: <3ED5923C.20EA243E@mimos.my> Message-ID: <3ED63C00.1010504@iai.uni-sb.de> You can find a description of all the used methods in: http://stp.ling.uu.se/~ebbag/somers.pdf or an updated version of this paper in: Michael Carl & Andy Way, Recent Advances in Example-based Machine Translation, Text, Speech and Language Technology Series, Kluwer Academic Publishers (to appear). there was a workshop on EBMT which includes many references: http://www.compapp.dcu.ie/~away/EBMT.html don't miss your compatriot's work: http://utmk.cs.usm.my/content/staff/mosleh.htm There are also more statistics-based approaches, for instance at ISI : http://www.isi.edu/~marcu/papers/transmem-acl01.pdf or at RALI (Michel Simard & Philippe Langlais) http://www.iro.umontreal.ca/~felipe/Papers/amta-2002.pdf good luck, Michael Dr Normaziah Abdul Aziz wrote: >Dear MT list members, > >Recently, there were active discussions and exchange of ideas on >Example-based Machine Translation (EBMT) issues in this mailing list. I >found that was interesting and useful discussions. > >In relation to EBMT, I am looking into various methods of selecting the >best or closest similar sentences from examples stored in the parallel >text database. >Has any of you worked on this and/or could give some pointers on such >selection methodology that is acceptablely functioning for English >language. > >Thanks, > >Dr. Normaziah >MIMOS >Malaysia > > > > From toby@ccl.umist.ac.uk Fri May 30 10:54:50 2003 From: toby@ccl.umist.ac.uk (Toby Peers) Date: Fri, 30 May 2003 10:54:50 +0100 Subject: [MT-list] Is this hidden text technique original? In-Reply-To: <3ED38D2D.A3F4BF73@mitre.org> References: <1051010433.3ea5258175dd6@webmail1.umist.ac.uk> <3ED38D2D.A3F4BF73@mitre.org> Message-ID: <1054288490.3ed72a6a967b3@webmail1.umist.ac.uk> Quoting Florence Reeder : > Toby, > > The hiding of text for forcing MT in HTML is unique, although > the tagging of items to prevent translation is not. The interesting > bit would be to see how systran processes this... Do they ignore > the "hidden" text or do they process it as a separate translation > unit? > > flo > The following is the the original page at the link i sent 1) 2) 3) 4)Test of invisible words in MT 5) 6) 7) 8[invisible]) The 9[visible]) open 10[invisible]) door was red. The 11[visible])mail 12[invisible] ) was light. 13)
14)

15 [visible])open mail 16)

17)

18 [visible])The open door was red. The mail was light. 19) This is the resulting systran output 1b) 2b) 3b) 4b)Test der unsichtbaren Wörter in der M.Ü. 5b) 6b) 7b) 8b [invisible] 9b [visible])geöffnete 10b [invisible]) Tür war rot. 11b [visible])Post 12b [invisible]) war Licht. 13b)
14b)

15b [visible])öffnen Sie Post 16b)

17b)

18b [visible])Die geöffnete Tür war rot. Die Post war Licht. 19b) Both the invisible and visible text is translated and this is exactly the reason why the effect works. By treating the sentences containing visible text and invisible text exactly as it would a normal purely visible sentence, Systran allows us to use the hidden context of the invisible text to determine its choice of translation alternatives. Note in 8) ,8b) and 10) and 10b) Systran doesn't translate 'the' to 'der/die or das'. I wonder if using complete sentences is necessary and whether single invisible modifiers might be more practical, e.g. (NP) --> NP The table --> the table leg The table --> the table entry The cycle --> the cycle frame The cycle --> the cycle duration AP --> AP strong --> strong smelling strong --> strong rugged watch --> wrist watch watch --> night watch Verbs may be a little more problematic as I found thinking up suitable adverbs much harder than the above noun and adjective modifiers VPbar? --> VPbar? plays --> plays musically plays --> plays fairly opens --> opens widely opens --> opens ???? What word would imply the opening of a bank account? I haven't tested this yet maybe I should Toby ************************************************************* Toby Peers Centre for Computational Linguistics UMIST PO Box 88 Tel: +44 161 200 3077 Manchester M60 1QD Fax: +44 161 200 3091 UK toby@ccl.umist.ac.uk ************************************************************* From toby@ccl.umist.ac.uk Fri May 30 12:39:11 2003 From: toby@ccl.umist.ac.uk (Toby Peers) Date: Fri, 30 May 2003 12:39:11 +0100 Subject: [MT-list] Is this hidden text technique original? In-Reply-To: <00f501c3269b$18be0ac0$12941e3e@ibm55h579k> References: <1051010433.3ea5258175dd6@webmail1.umist.ac.uk> <3ED38D2D.A3F4BF73@mitre.org> <1054288490.3ed72a6a967b3@webmail1.umist.ac.uk> <00f501c3269b$18be0ac0$12941e3e@ibm55h579k> Message-ID: <1054294751.3ed742dfcbd62@webmail1.umist.ac.uk> Quoting Roger Harris : > 2003 May 30. London, U.K. > ---------------------------------------------------------------------------- > --- > > > And it has been said .. > >>> The hiding of text for forcing MT in HTML is unique, although > >>> the tagging of items to prevent translation is not. The interesting > >>> bit would be to see how systran processes this... > > Systran can hardly respond correctly to an HTML document if it has not been > programmed to recognise one. Roger, The RAW html code was not being translated. For example, by being pasted into the text box at http://www.google.co.uk/language_tools?hl=en rather, I used the translate a web page option on the above page. Here you can paste the URL of a complete web page, which systran then translates as a entire web page. they have obviously developed some server side code to scrape the source web page and separate the textual content from the structural markup, process the content, leave the markup alone and then reassemble it all back into a functioning web page. The URL I pasted into the google translation service was http://mull.ccl.umist.ac.uk/staff/toby/Partial_Translation_in_CALL/testing _systran/invisible_text.html The code I wrote into my mail of this morning was a) the source code of this page and b) the resulting source code of google/systrans' English--> German processed web page output. So systran was not being forced to translate raw html. Toby ************************************************************* Toby Peers Centre for Computational Linguistics UMIST PO Box 88 Tel: +44 161 200 3077 Manchester M60 1QD Fax: +44 161 200 3091 UK toby@ccl.umist.ac.uk ************************************************************* From Roger Harris"

Program of the Workshop = "Natural Language Processing and Multilingualism"
http://tilt.elibel.tm.fr/events/conferences/workshops/nlp-multil= inguism/

in conjunction with TALN 2003 = Conference
http://www.sci= ences.univ-nantes.fr/irin/taln2003/

Batz-sur-Mer, France
June 11-14, 2003

Organizers of the workshop = :
Malek Boualem = (malek.boualem@rd.francetelecom.com)
Emilie Guimier De Neef = (emilie.guimierdeneef@rd.francetelecom.com)

------_=_NextPart_001_01C32F69.42C8D41E-- From debe@comp.leeds.ac.uk Fri Jun 13 13:00:20 2003 From: debe@comp.leeds.ac.uk (D Elliott) Date: Fri, 13 Jun 2003 13:00:20 +0100 (BST) Subject: [MT-List] Summary: Parallel texts for MT evaluation Message-ID: Dear all, Thanks to everyone who responded to my request for parallel texts with good quality human translations, suitable for my MT evaluation research. Here is a summary of resources available from the web: INTERSECT corpus FRENCH-ENGLISH: Le Monde, instructions for domestic appliances, technical and academic texts and others GERMAN-ENGLISH Company home pages, news items, EU documents and more http://www.brighton.ac.uk/edusport/languages/html/intersect.html Thanks to Professor Raphael Salkie, University of Brighton, UK Proceedings of the European Parliament MANY EUROPEAN LANGUAGES INTO ENGLISH http://www.isi.edu/~koehn/publications/europarl/ Thanks to Susana Sotelo Docío, Universidade de Santiago de Compostela OPUS corpus ENGLISH SOURCE TEXTS translated into French, Spanish, Swedish, German, and Japanese. Jörg Tiedemann and Lars Nygaaard compiled the documentation of the office package OpenOffice[1] and the PHP[2] manual. The resulting corpus is OPUS - an open source parallel corpus. http://logos.uio.no/opus/ [1] http://www.openoffice.org [2] http://www.php.net Thanks to Susana Sotelo Docío, Universidade de Santiago de Compostela UN declarations of human rights Many languages http://www.unhchr.ch/udhr/index.htm Thanks to Paul McNamee, Johns Hopkins University and Ella Earp-Lynch, SpeechWorks International Centre for Disease Control (USA) Chinese, French, Japanese, Spanish info on SARS and many other medical topics http://www.cdc.gov/ http://www.cdc.gov/ncidod/sars/languages.htm Thanks to Paul McNamee, Johns Hopkins University Debian free software community: Technical translations http://www.debian.org/international/ Thanks to Paul McNamee, Johns Hopkins University Official journal of the EU Freely downloadable European legislation in many languages http://europa.eu.int Thanks to Paul McNamee, Johns Hopkins University, Terence Lewis (Language Engineer) and Koen.Kerremans Public registry of the Council of the EU PDF files in various languages. Translations indicate the source language. http://register.consilium.eu.int/ Thanks to John Beaven COMPARA corpus English-Portuguese/Portuguese-English http://www.linguateca.pt/COMPARA/ Thanks to Dr Ana Frankenberg-Garcia,Instituto Superior de Línguas e Administraçăo, Lisboa, Portugal The Universal Declaration of Human Rights UNESCO's website also has most documents available translated into Spanish, French and frequently into Russian, Chinese and Arabic French Foreign Ministry's magazine - Label France: French into various languages http://www.france.diplomatie.fr/label_france/index.html Thanks to Jeremy Whistle, University College Northampton ELRA newsletter In French and English www.elda.fr Thanks to Jeff Allen Multilingual articles: English version: http://www.multilingual.com/allen51.htm French translation: http://www.editionscle.com/bol/presse/article1/allen-mltc51-fr.htm English version: http://www.multilingual.com/allen53.htm French translation: http://www.editionscle.com/bol/presse/article2/allen-mltc53-fr.htm Thanks to Jeff Allen Haitian Creole version: http://hometown.aol.com/mit2haiti/JA-HC-kr.htm English version: http://hometown.aol.com/mit2haiti/JA-HC-eng.htm Thanks to Jeff Allen MIT2 website Marilyn Mason Bio & Publication List: http://hometown.aol.com/marilinc/Index3.html Creole Links Page: http://hometown.aol.com/mit2haiti/Index4.html The Creole Clearinghouse: http://hometown.aol.com/CreoleCH/Index6.html Thanks to Jeff Allen -- *************************************************** Debbie Elliott Computer Vision and Language Research Group, School of Computing, University of Leeds, Leeds LS2 9JT United Kingdom. Tel: 0113 3436818 Email: debe@comp.leeds.ac.uk Website (to be expanded): http://www.comp.leeds.ac.uk/cgi-bin/sis/ext/rs_pub.cgi/debe.html?cmd=displayrs *************************************************** From abcs@online.no Fri Jun 13 14:36:56 2003 From: abcs@online.no (=?ISO-8859-1?Q?=D8yvind?= Str=?ISO-8859-1?Q?=F8m?=) Date: Fri, 13 Jun 2003 15:36:56 +0200 Subject: [MT-List] Universal Translator 2000 Message-ID: <3EEC00B7@epostleser.online.no> Hello New to the list, I would like ask if someone has any experience with the product Universal Translator 2000. On paper it looks promising. How is the quality of the translation? Can it communicate with other applications, without interaction? (ActiveX, COM ??) Who owns the product? (After Languageforce went down) How can one get in contact with them? How is the quality of support? I would also very much like to receive recommendations of other MT-packages, preferably omnidirectional ones, beiing able to communicate with other applications (without interaction) Thanks a lot =D8yvind Str=F8m From cb@lim.nl Mon Jun 16 07:00:10 2003 From: cb@lim.nl (Colin Brace) Date: Mon, 16 Jun 2003 08:00:10 +0200 Subject: [MT-List] Fwd: MT Summit: registration now open Message-ID: <20030616060107.KECM17933.amsfep14-int.chello.nl@[62.108.30.75]> ------------------------------------------------------------------------- The following message is forwarded to you by Colin Brace ------------------------------------------------------------------------- >Message-Id: >Date: Sat, 14 Jun 2003 11:53:09 -0700 >From: Eduard Hovy >To: webmaster@eamt.org >Subject: MT Summit: registration now open Announcement International Machine Translation Summit IX www.mt-summit.org September 23-27, New Orleans, USA **** Registration is now open: www.mt-summit.org **** The ninth Machine Translation Summit, organized by the International Association for Machine Translation (IAMT) and hosted by the Association for Machine Translation in the Americas (AMTA), will be held in New Orleans, Louisiana, from 23 to 27 September 2003. The MT Summit IX will feature a comprehensive program that includes research papers, reports on users' experiences, invited talks, discussions of policy issues, panels, exhibits, 4 tutorials, and 4 workshops. We define machine translation in the broadest possible sense, to include not just fully automatic MT but tools for translation support and multilingual text processing as well. Speech-to-speech translation is one of the themes of the conference this year. A lively social agenda will include a reception and a surprise banquet that promises a very enjoyable evening. The conference hotel offers a stunning and elegant setting for a conference and is within 5 minutes' walking distance from the historic French Quarter, birthplace of jazz. Other accommodation is within two blocks. We invite all those with an interest in translation automation--researchers, developers, translation service providers, users, and managers--to participate in the conference. ----------------------------------------------------- -- End of forwarded message ----------------------------------------------------- -- Colin Brace | cb@lim.nl Amsterdam http://www.lim.nl From WJHutchins@compuserve.com Mon Jun 16 09:59:39 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Mon, 16 Jun 2003 04:59:39 -0400 Subject: [MT-List] Universal Translator 2000 Message-ID: <200306160459_MC3-1-3D6D-5877@compuserve.com> Dear Mr Str=F8m, > New to the list, I would like ask if someone has any experience with th= e = >product Universal Translator 2000. On paper it looks promising. I obtained a copy a few years ago -- in order to see how bad it was. >How is the quality of the translation? The quality is extremely poor, and I would not recommend it. >Can it communicate with other applications, without interaction? (Active= X, COM = ??) Perhaps, I have not tested it .. >Who owns the product? (After Languageforce went down) Nobody as far as I know. I was told that a company called Glossa was continuing to develop and/or sell it, but I have been unable to find them on the Internet. >How can one get in contact with them? Don't know >How is the quality of support? It was poor before the company ceased operation >I would also very much like to receive recommendations of other MT-packages, >preferably omnidirectional ones, beiing able to communicate with other >applications (without interaction) Which langauegs are you interesetd in? For a full listing of current commercial systems see the "Compendium of translation software" on the EAMT website. With regards, John Hutchins 16 June =D8yvind Str=F8m < From jack@kanji.org Tue Jun 17 01:07:02 2003 From: jack@kanji.org (Jack Halpern) Date: Mon, 16 Jun 2003 19:07:02 -0500 Subject: [MT-List] Universal Translator 2000 In-Reply-To: <200306160459_MC3-1-3D6D-5877@compuserve.com> References: <200306160459_MC3-1-3D6D-5877@compuserve.com> Message-ID: <200306170007.AA04840@mail.kanji.org> John Hutchins wrote... >Dear Mr Str$B".(B, > >> New to the list, I would like ask if someone has any experience with the >>product Universal Translator 2000. On paper it looks promising. > I obtained a copy a few years ago -- in order to see >how bad it was. >>How is the quality of the translation? > The quality is extremely poor, and I would not recommend it. I totally agree. And the software is difficult to install and did not work after months of trying to get technical support. >>Can it communicate with other applications, without interaction? (ActiveX, >COM >??) > Perhaps, I have not tested it .. > >>Who owns the product? (After Languageforce went down) > Nobody as far as I know. I was told that a company called Glossa >was continuing to develop and/or sell it, but I have been unable to find >them on the Internet. >>How can one get in contact with them? I know the original people who were running the company. The CEO was Ian Simpson , but the address may not be current. > Don't know >>How is the quality of support? > It was poor before the company ceased operation >>I would also very much like to receive recommendations of other >MT-packages, >>preferably omnidirectional ones, beiing able to communicate with other >>applications (without interaction) > Which langauegs are you interesetd in? > For a full listing of current commercial systems see the >"Compendium >of translation software" on the EAMT website. > >With regards, >John Hutchins >16 June >$B%j(Byvind Str$B".(B >< > > >-- > For MT-List info, see http://www.eamt.org/mt-list.html > Regards, Jack Halpern President, The CJK Dictionary Institute, Inc. http://www.cjk.org Phone: +81-48-473-3508 From nadamides@aslib.com" TRANSLATING AND THE COMPUTER 25 CONFERENCE, 20-21 November 2003, London This conference is one of the few international events which focuses on the user aspects of translation software and as such has been particularly beneficial to a very wide audience including translators, business managers, researchers and language experts. The event, which is organised by Aslib, has the support of EAMT, IAMT, BCS - Natural Language Group, Institute of Linguists. Once again, this year the conference will address the latest developments in translation (and translation-related) software. It will address the needs of the following conference attendees: industry public administration agencies freelancers development This call for papers invites abstracts of papers to be presented at the conference. The papers (and the presentations) should focus on the user aspects of translation or translation-related software rather than on theoretical issues. Presentations accompanied by demonstrations are espe cially welcome. TOPICS The range of topics includes (but is not limited to) use of MT systems machine-aided translation and translation aids controlled languages and their use in MT speech translation terminology localisation multilingual document management/workflow case studies of technology-based solutions the Internet and translation aids/services the value of "free" versus "charging" services/sites on the Internet SUBMISSION GUIDELINES Authors are required to submit an abstract of a MINIMUM of 500 words of the paper they would like to present, together with an outline of the structure of the paper and short BIOGRAPHY. Abstracts should be sent by POST or EMAIL before 25 June 2003 (extended deadline) to: Nicole Adamides, Conference Organiser Aslib, The Association for Information Management Temple Chambers, 3-7 Temple Avenue, London, EC4Y 0HP Tel: +44(0) 20 7583 8900 Fax: +44 (0) 20 7583 8401 Email: nadmides@aslib.com The abstracts will be considered by the Programme Chairs, namely: Daniel Grasmick and Chris Pyne, SAP; Professor Ruslan Mitkov, University of Wolverhampton and Olaf-Michael Stefanov, United Nations. The authors of abstracts will be notified of acceptance or rejection of their submissions by 1 August 2003. The full length versions of the accepted papers (authors will be provided with detailed camera-ready copy guidelines) will be included in the conference proceedings and must be submitted by 10th October 2003. NICOLE ADAMIDES, Training Aslib/IMI, Temple Chambers, 3-7 Temple Avenue, London, EC4Y 0HP Tel: +44 (0)20 7583 8900 Fax: +44 (0)20 7583 8401 www.aslib.com Email: nadamides@aslib.com From hanne@cst.dk Mon Jun 16 11:07:54 2003 From: hanne@cst.dk (Hanne =?iso-8859-1?Q?Fers=F8e?=) Date: Mon, 16 Jun 2003 12:07:54 +0200 Subject: [MT-List] Universal Translator 2000 References: <200306160459_MC3-1-3D6D-5877@compuserve.com> Message-ID: <3EED96FA.EF66F957@cst.dk> Maybe I can be of help here. At CST we receive electronic marketing material from a Nordic reseller (or maybe they are also producers) from time to time. Currently they are called st@rgate, previously they were called Team Work Nordic. They both seem to be or have been located in Norway. The e-mail address sending the latest information (received on June 10th this year) was stargate@bef.no. The marketing letter offers a reduced price 498, - kr. (ordinary price 2,990 kr.). There is a reference to http://www.stargate-as.com where you may order your copy. We never conducted a systematic test, but we have tried out a few things, and I can safely say that our impression is very much the same as John Hutcins's. Yours, Hanne Fersře John Hutchins wrote: > > Dear Mr Strřm, > > > New to the list, I would like ask if someone has any experience with the > >product Universal Translator 2000. On paper it looks promising. > I obtained a copy a few years ago -- in order to see > how bad it was. > >How is the quality of the translation? > The quality is extremely poor, and I would not recommend it. > > >Can it communicate with other applications, without interaction? (ActiveX, > COM > ??) > Perhaps, I have not tested it .. > > >Who owns the product? (After Languageforce went down) > Nobody as far as I know. I was told that a company called Glossa > was continuing to develop and/or sell it, but I have been unable to find > them on the Internet. > >How can one get in contact with them? > Don't know > >How is the quality of support? > It was poor before the company ceased operation > >I would also very much like to receive recommendations of other > MT-packages, > >preferably omnidirectional ones, beiing able to communicate with other > >applications (without interaction) > Which langauegs are you interesetd in? > For a full listing of current commercial systems see the > "Compendium > of translation software" on the EAMT website. > > With regards, > John Hutchins > 16 June > Řyvind Strřm > < > > -- > For MT-List info, see http://www.eamt.org/mt-list.html -- Hanne Fersře Souschef /Deputy Manager Center for Sprogteknologi Njalsgade 80, DK-2300 Křbenhavn S Tel: +45 35 32 90 73 Fax: +45 35 32 90 89 From ling98@canada.com Wed Jun 18 00:45:32 2003 From: ling98@canada.com (Michael Blekhman) Date: Tue, 17 Jun 2003 19:45:32 -0400 Subject: [MT-List] International Journal for Translation References: <8C19E3FBB6467846AFF97E366D34CB60010AFEF5@lanmhs20.rd.francetelecom.fr> Message-ID: <011301c3352a$8a7926c0$6400a8c0@michaelb> Program of the Workshop "NLP and Multilingualism"Dear colleagues, Sorry if you have received this message by mistake or more than once. I am happy to inform you that the Machine Translation January-June 2003 issue of the International Journal of Translation is out of press. I am now in the process of editing the July-December 2003 issue, which will combine articles on machine translation and other issues of translation and linguistics. The deadline for that issue is mid-August 2003, so please feel free to send me your contributions. Besides, I invite you to participate in the 2004 issue on machine translation, the deadline being mid-January 2004. The E-mail addresses to send your articles to: ling98@canada.com or ling98@videotron.ca The requirements are very simple to meet: - up to 24 pages in RTF or MS Word; - font: Times New Roman; - 12 pts. Looking forward to hearing from you! Regards, Michael Blekhman Lingvistica '98 Inc. Montreal, Canada From ling98@canada.com Wed Jun 18 00:45:32 2003 From: ling98@canada.com (Michael Blekhman) Date: Tue, 17 Jun 2003 19:45:32 -0400 Subject: [MT-List] International Journal for Translation References: <8C19E3FBB6467846AFF97E366D34CB60010AFEF5@lanmhs20.rd.francetelecom.fr> Message-ID: <011401c3352a$8d41bca0$6400a8c0@michaelb> Program of the Workshop "NLP and Multilingualism"Dear colleagues, Sorry if you have received this message by mistake or more than once. I am happy to inform you that the Machine Translation January-June 2003 issue of the International Journal of Translation is out of press. I am now in the process of editing the July-December 2003 issue, which will combine articles on machine translation and other issues of translation and linguistics. The deadline for that issue is mid-August 2003, so please feel free to send me your contributions. Besides, I invite you to participate in the 2004 issue on machine translation, the deadline being mid-January 2004. The E-mail addresses to send your articles to: ling98@canada.com or ling98@videotron.ca The requirements are very simple to meet: - up to 24 pages in RTF or MS Word; - font: Times New Roman; - 12 pts. Looking forward to hearing from you! Regards, Michael Blekhman Lingvistica '98 Inc. Montreal, Canada From WJHutchins@compuserve.com Fri Jun 20 10:09:31 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Fri, 20 Jun 2003 05:09:31 -0400 Subject: [MT-List] Universal Translator 2000 Message-ID: <200306200509_MC3-1-3E28-90E4@compuserve.com> Dear Hanne Fers=F8e and MT-list, Thank you for alerting me to this. I have looked at the Stargate website and found that they are selling Universal Translator at $249.95 (it seems to be their only product). From the look of the package and the languages covered it would seem to be identical to the system sold previously by LanguageForce. There is rather a steep price increase -- from $39 (December 2000) to $249 ! but I doubt that the content is any different. John Hutchins 19 June =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > = Maybe I can be of help here. At CST we receive electronic marketing material from a Nordic reseller (or maybe they are also producers) from time to time. Currently they are called st@rgate, previously they were called Team Work Nordic. They both seem to be or have been located in Norway. The e-mail address sending the latest information (received on June 10th this year) was stargate@bef.no. The marketing letter offers a reduced price 498, - kr. (ordinary price 2,990 kr.). There is a reference to http://www.stargate-as.com where you may order your copy. We never conducted a systematic test, but we have tried out a few things, and I can safely say that our impression is very much the same as John Hutcins's. Yours, Hanne Fers=F8e< From lorena.guerra2@mail.dcu.ie Tue Jun 24 11:16:21 2003 From: lorena.guerra2@mail.dcu.ie (lorena.guerra2@mail.dcu.ie) Date: Tue, 24 Jun 2003 11:16:21 +0100 Subject: [MT-List] MT is worthwhile for...? Message-ID: <3EA4FB980004662B@hawk.dcu.ie> Dear all, I was wondering if someone knows of any specific citations in publication= s or any article(s) that discourage MT for High Quality translation efforts= , and/or those that specifically promote MT systems only for dissemination purposes using technical manuals. Regards, Lorena Guerra, Postgraduate student at DCU- Ireland. Information needed for my dissertation in Translation Technology. From harold.somers@umist.ac.uk Tue Jun 24 10:26:22 2003 From: harold.somers@umist.ac.uk (Harold Somers) Date: Tue, 24 Jun 2003 10:26:22 +0100 Subject: [MT-List] MT is worthwhile for...? In-Reply-To: <3EA4FB980004662B@hawk.dcu.ie> Message-ID: > I was wondering if someone knows of any specific citations in > publications or any article(s) that discourage MT for High Quality > translation efforts, and/or those that specifically promote MT systems > only for dissemination purposes using technical manuals. > Well almost any (respectable) article on MT will say something along those lines. With apologies for shameless self-promotion, the point is made repeatedly by me and other contributors to H. Somers (ed.) Computers and Translation: A Translator=92s Guide (Benjamins Translation Library 35), Amsterdam (2003): John Benjamins, now available at the awesome price of =80115 (sorry - beyond my control!) From Christian.Boitet@imag.fr Tue Jun 24 11:47:36 2003 From: Christian.Boitet@imag.fr (Christian Boitet) Date: Tue, 24 Jun 2003 12:47:36 +0200 Subject: [MT-List] MT is worthwhile for...? In-Reply-To: <3EA4FB980004662B@hawk.dcu.ie> References: <3EA4FB980004662B@hawk.dcu.ie> Message-ID: Dear Lorena, 24/6/03 At 11:16 +0100 24/06/03, lorena.guerra2@mail.dcu.ie wrote: >Dear all, > >I was wondering if someone knows of any specific citations in publications >or any article(s) that discourage MT for High Quality translation efforts, >and/or those that specifically promote MT systems only for dissemination >purposes using >technical manuals. There are such articles with these 2 points of view. But both are too naive. HQMT can be achieved and beat HT in some contexts. That means that postedition of raw MT costs less thant postedition of HT. revision can be counted in minutes/page (20 after "standard" HT) or number of editing operations per 100 words (20-25 after stantard HT), if counting insertions and suppressions of entire words (replacement costs 2, exchange 2, swapping around a segment 4). See reports on Meteo (3% revision), CATALYST (CMU-Caterpillar), ALT/Flash, even METAL (Slocum 1984). HQMT can also be achieved with human intervention during or after analysis, with a view to multitarget translation. See Herv=E9 Blanchon's thesis on all such "interactive systems", plus a paper in MT (1994-95). You may again evaluate quality by the human time taken to obtain N polished translations (maybe 15 minutes/page of interactive disambiguation and 5mn/p of postedition instead of 20, for each target language). To advocate MT only for dissemination of technical manuals is plain silly when considering the millions of pages translated by MT web servers everyday on request of internauts wanting to access to information. Please keep in mind the figures from the evaluation of the GAT (Georgetown Automatic Translation) system at Euratom, Ispra, around 1972: linguistic quality 1/10, usefulness 9/10. They still apply=8A >Regards, > >Lorena Guerra, >Postgraduate student at DCU- Ireland. >Information needed for my dissertation in Translation Technology. Best regards, Ch.Boitet From ling98@canada.com Wed Jun 25 18:50:21 2003 From: ling98@canada.com (Michael Blekhman) Date: Wed, 25 Jun 2003 13:50:21 -0400 Subject: [MT-List] International Journal of Translation Message-ID: <00a001c33b42$3f8b9b20$6400a8c0@michaelb> Dear colleagues, Sorry if you have received this message by mistake or more than once. I am happy to inform you that the Machine Translation January-June 2003 issue of the International Journal of Translation is out of press. I am now in the process of editing the July-December 2003 issue, which will combine articles on machine translation and other issues of translation and linguistics. The deadline for that issue is mid-August 2003, so please feel free to send me your contributions. Besides, I invite you to participate in the 2004 issue on machine translation, the deadline being mid-January 2004. The E-mail addresses to send your articles to: ling98@canada.com or ling98@videotron.ca The requirements are very simple to meet: - up to 24 pages in RTF or MS Word; - font: Times New Roman; - 12 pts. Looking forward to hearing from you! Regards, Michael Blekhman Lingvistica '98 Inc. Montreal, Canada From Andy.Way" Two funded Ph.D. scholarships in Machine Translation 2003/2006 The School of Computing and the National Centre for Language Technology studentships for an Enterprise Ireland funded Basic Research project on Machine Translation. The project starts October 2003, and funding is for three years at a rate of 12,000+ euro p/a and fees. Our ideal candidates would have a good background in Computing, Computational Linguistics and Natural Language Processing; programming experience in one or more of Java, Perl, C++, Prolog, Lisp etc, and a strong interest in and motivation for research. The research involves probabilistic (PCFG etc.) and finite state-based corpus analysis techniques. For informal inquiries and further details, please contact josef@computing.dcu.ie, away@computing.dcu.ie by Friday July 11th, 2003. Josef van Genabith and Andy Way School of Computing National Centre for Language Technology Dublin City University, Dublin 9, Ireland http://www.computing.dcu.ie/research/nclt/ From lorena.guerra2@mail.dcu.ie Mon Jul 7 19:14:19 2003 From: lorena.guerra2@mail.dcu.ie (lorena.guerra2@mail.dcu.ie) Date: Mon, 7 Jul 2003 19:14:19 +0100 Subject: [MT-List] Who should be doing MT post-editing? Message-ID: <3EFBE21800008CC1@hawk.dcu.ie> Dear all, I would like to ask you, from your experience point of view: Who(do you think) should be doing Machine translation Post-editing? Should this activity be performed by (junior/senior) translators, reviser= s, non-linguistics, trained specialists...? Why? Regards, Lorena Guerra Postgraduate student at DCU- Ireland Information needed for my dissertation in Translation Technology From jeff.allen@free.fr Tue Jul 8 17:59:15 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Tue, 8 Jul 2003 18:59:15 +0200 Subject: [MT-List] Who should be doing MT post-editing? Message-ID: <1057683555.3f0af8638d0ea@impt1-1.free.fr> Quoting posting to MT-list by lorena.guerra2@mail.dcu.ie > I would like to ask you, from your experience point of view: > Who(do you think) should be doing Machine translation Post-editing? > Should this activity be performed by (junior/senior) translators, revisers, > non-linguistics, trained specialists...? Why? > Regards, > Lorena Guerra > Postgraduate student at DCU- Ireland > Information needed for my dissertation in Translation Technology * Students: Mention of students doing postediting either in their native language or not: chap 17. Machine translation in the classroom Harold Somers section 2.2 post-editing (p 322) see http://www.benjamins.com/cgi-bin/t_bookview.cgi?bookid=BTL_35 Also mentioned in an example in same chapter, section 4.2 Assessing MT for assimilation (p 331) The example shows an online translation of a web page and proposes the idea of an exercise to post-edit the page in target language without seeing the original source text or knowing the source language. I usually discourage this. * marketing and product specialists: marketing product line and service specialists can do postediting See: ALLEN, Jeffrey. 2001. Postediting: an integrated part of a translation software program. In Language International magazine, April 2001, pp. 26-29. * Human Resources: Training on MT postediting of some tech doc writers and Human Resources specialists took place today. The Human Resources people write up external company communication (ie job announcements) and company internal communication in at least 2 languages. * Bilingual secretaries: Last month a couple of bilingual secretaries were trained on use of MT software for multilingual documentation that they communicate to their groups and associations that partner with their association. * levels of postediting: The types of posteditors really depends on the type of postediting that is conducted (full, minimal, rapid) as described in: chap 16. Post-editing Jeffrey Allen (pp 297-319) see http://www.benjamins.com/cgi-bin/t_bookview.cgi?bookid=BTL_35 I believe that Postediting can be taught to anyone (with bilingual skills) involved in the field of communication of information. * Native tongue postediting or not? Postediting into ones native tongue or not can be debated depending on the level of postediting that is assigned and the quality of the translation that is expected. Most everything I have read about postediting has been from the perspective of training translators to conduct postediting tasks. >From my experience: * experienced translators and translation revisers/editors: In the event that the use of minimal postediting results in the creation of a sublanguage in the target language, it is sometimes difficult for experienced translators and translation editors to accept translating and work with with a level of quality that is lower than what they have done for years. * junior translators: Training junior translators has the benefit of having to spend less work helping untrain people on what they have done for years. They are often easier to convince that the postediting task is just one method of translating rather than simply a complete rearrangement of a translation process cycle, or simply a substandard level of translation quality. Regards, Jeff From jeff.allen@free.fr Wed Jul 9 13:06:25 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Wed, 9 Jul 2003 14:06:25 +0200 Subject: [MT-List] MT is worthwhile for...? Message-ID: <1057752385.3f0c05414e314@impt2-2.free.fr> Lorena Guerra wrote: >> I was wondering if someone knows of any specific citations in >> publications or any article(s) that discourage MT for High Quality >> translation efforts, and/or those that specifically promote MT systems >> only for dissemination purposes using technical manuals. Harold Somers replied: >Well almost any (respectable) article on MT will say something >along those lines. With apologies for shameless self-promotion, the >point is made repeatedly by me and other contributors to H. >Somers (ed.) >Computers and Translation: A Translator’s Guide (Benjamins >Translation Library 35), Amsterdam (2003): John Benjamins, I've actually started saying the opposite in recent articles (ie, encouraging the use of MT for high-quality translation) ONLY for the scenario that an MT system with a reasonable amount of MT postediting features is available to the posteditor. Definitely NOT using online web portals that are aimed for content gisting MT usage. See: ALLEN, Jeffrey. 2001. Postediting: an integrated part of a translation software program. In Language International magazine, 13:2, April 2001, pp. 26- 29. ALLEN, Jeffrey. 2002. Review of "Repairing Texts: Empirical Investigations of Machine Translation Post-Editing Processes". (KRINGS Hans, edited by Geoffrey KOBY. 2001. Ohio: Kent State University Press. (abstract and ordering information at http://bookmasters.com/ksu-press/ksu071.htm ) In Multilingual Computing and Technology magazine. Number 46. March 2002. Pp. 27-29. Available at: http://www.multilingual.com/allen46.htm ALLEN, Jeffrey. 2003. Panel talk on Controlled Language and Machine Translation Implementation and Use. Presented at the European Association for Machine Translation and Controlled Language Applications Workshop (EAMT/CLAW 2003). As mentioned during that panel talk, and indicated in the powerpoint presentation, the following article is living proof of MT for high-quality publication. English orginal containing just under 6,000 words was minimally to mid-level-postedited (in between minimal and full postediting) in French in 6 hours (ie 1000 words per hour on average). ALLEN, Jeff. 2002. The Bible as a Resource for Translation Software: A proposal for MT development using an untapped language resource database. In Multilingual Computing and Technology magazine. Number 51, Vol. 13, Issue 7. October/November 2002. Pp. 40-45. English version: http://www.multilingual.com/allen51.htm French translation: La Bible comme Ressource pour les Logiciels de Traduction: Une proposition de développement des systčmes de traduction automatique (TA) en utilisant une ressource linguistique inexploitée http://www.editionscle.com/bol/presse/article1/allen-mltc51-fr.htm Steps conducted: minimal to mid-level postediting, then translation review (by another person) and then final review by the original posteditor. Electronic copies of the French version were made at each hour of the postediting stage, as well as at the end of the postediting stage, the first review stage, and the final review stage. At this rate, a full work day could produce 8000-9000 postedited words, when the posteditor knows the subject very well. Note: My current survey on translation speed amoung many organizations and independent translators indicates an average translation speed of 2000-3000 words per work day. Note: An average typist (30 words per minute) doing full sprint typing for 8 hours (if the person could survive such a duration) could theoretically reproduce from softcopy paper to electronic format a total 14,400 words. However, most typists cannot keep such a typing rate for 8 solid hours. So, postediting can be equal to or better than the typing speed of an average typist. However, we must remember that translation includes cognitive effort beyond simply reproducing what is already written down. In this study, the initial postediting within an MT postediting + review cycle can be performed at nearly 3 times the average translation speed based on the specific conditions below: * the posteditor knows the subject very well. * posteditor is near-native (but not native) speaker of the target language * English to French translation pair * the posteditor is an expert user of the MT tool in question * no dictionary building was conducted for this specific experiment * first step of minimal to mid-level postediting followed then by 2 additional translation review steps to produce a high-quality target language document. A couple of upcoming articles will describe and build upon these results in more detail. Best, Jeff From kth53827@cs.usm.my Mon Jul 14 10:55:17 2003 From: kth53827@cs.usm.my (tuan hwa) Date: Mon, 14 Jul 2003 17:55:17 +0800 Subject: [MT-List] Where can I download an articles from WWW References: <1057683555.3f0af8638d0ea@impt1-1.free.fr> Message-ID: <3F127E05.000001.33460@UTM-TH> --------------Boundary-00=_58E0QL80000000000000 Content-Type: Multipart/Alternative; boundary="------------Boundary-00=_58E0LVC0000000000000" --------------Boundary-00=_58E0LVC0000000000000 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear all, =0D I am a student who doing a research in EBMT. However, until now, I cannot find an article as belows: =0D =0D Nagao, M. (1984) 'A Framework of a Mechanical Translation between Japanes= e and English by Analogy Principle', in A. Elithorn & R. Banerji(eds), eds.= , Artificial and Human Intelligence, Nato Publication.=0D =0D Hope that you can help me, thank you.=0D Tuan Hwa --------------Boundary-00=_58E0LVC0000000000000 Content-Type: Text/HTML; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Dear all,
I am a student who doing a research in EBMT. However, until now, I c= annot find an article as belows:
 
Nagao, M. (1984) 'A Framework of a Mechanical Translation between Ja= panese and English by Analogy Principle', in A. Elithorn & R. Banerji= (eds), eds., Artificial and Human Intelligence, Nato Publication.
 
Hope that you can help me, thank you.
Tuan Hwa
 
 
______________________= ______________________________
<= A href=3D"http://www.incredimail.com/redir.asp?ad_id=3D309&lang=3D9">= 3D""  IncrediMail - Email has= finally evolved - = Click Here --------------Boundary-00=_58E0LVC0000000000000-- --------------Boundary-00=_58E0QL80000000000000 Content-Type: image/gif; name="IMSTP.gif" Content-Transfer-Encoding: base64 Content-ID: R0lGODlhFAAPALMIAP9gAM9gAM8vAM9gL/+QL5AvAGAvAP9gL////wAAAAAAAAAAAAAAAAAAAAAA AAAAACH/C05FVFNDQVBFMi4wAwEAAAAh+QQJFAAIACwAAAAAFAAPAAAEVRDJSaudJuudrxlEKI6B URlCUYyjKpgYAKSgOBSCDEuGDKgrAtC3Q/R+hkPJEDgYCjpKr5A8WK9OaPFZwHoPqm3366VKyeRt E30tVVRscMHDqV/u+AgAIfkEBWQACAAsAAAAABQADwAABBIQyUmrvTjrzbv/YCiOZGmeaAQAIfkE CRQACAAsAgABABAADQAABEoQIUOrpXIOwrsPxiQUheeRAgUA49YNhbCqK1kS9grQhXGAhsDBUJgZ AL2Dcqkk7ogFpvRAokSn0p4PO6UIuUsQggSmFjKXdAgRAQAh+QQFCgAIACwAAAAAFAAPAAAEEhDJ Sau9OOvNu/9gKI5kaZ5oBAAh+QQJFAAIACwCAAEAEAANAAAEShAhQ6ulcg7Cuw/GJBSF55ECBQDj 1g2FsKorWRL2CtCFcYCGwMFQmBkAvYNyqSTuiAWm9ECiRKfSng87pQi5SxCCBKYWMpd0CBEBACH5 BAVkAAgALAAAAAAUAA8AAAQSEMlJq7046827/2AojmRpnmgEADs= --------------Boundary-00=_58E0QL80000000000000-- From jeff.allen@free.fr Mon Jul 14 20:02:36 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Mon, 14 Jul 2003 21:02:36 +0200 Subject: [MT-List] article: The Business Case for MT: The Breakthrough Is for Real Message-ID: <009001c34a92$9c6bd8a0$0e9c933e@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_01FC_01C34A4B.40F35F80 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi folks, A new article has appeared on the business case for MT. See: = http://www.lisa.org/archive_domain/newsletters/2003/2.6/vandermeer.html For more info, the author's contact info is: >Cross Language n.v. >Automating Multilingual Communications >www.crosslang.com >jaap.vandermeer@crosslang.com Best, Jeff --------------------------------- Jeff ALLEN Paris, FRANCE Mobile Tel: (+33) 6 86 06 87 59 e-mail: jeff.allen@free.fr =20 ------=_NextPart_000_01FC_01C34A4B.40F35F80 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi folks,
 
A new article has appeared on the business case for=20 MT.
See:  http://www.lisa.org/archive_domain/newsletters/2003/2.6/vandermee= r.html

For=20 more info, the author's contact info is:
 
>Cross Language n.v.
>Automating Multilingual=20 Communications
>www.crosslang.com
>jaap.vandermeer@crosslang.c= om
 
Best,
 
Jeff

---------------------------------
Jeff=20 ALLEN
Paris,  FRANCE
Mobile Tel: (+33) 6 86 06 87 = 59
e-mail: jeff.allen@free.fr
 
------=_NextPart_000_01FC_01C34A4B.40F35F80-- From michelewilliam.corradi@tin.it Mon Jul 21 08:03:07 2003 From: michelewilliam.corradi@tin.it (Mike Corradi) Date: Mon, 21 Jul 2003 09:03:07 +0200 Subject: [MT-List] Russian researchers Message-ID: <001401c34f56$23d2b8a0$fa4fb450@mike> This is a multi-part message in MIME format. ------=_NextPart_000_0011_01C34F66.E6BCD7A0 Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable Dear All,=20 I am currently working on my thesis at the University of Parma, it is on = MT with particular reference to Italian and Russian. At the moment I am = working on a short history section and I am looking for material on the = Russian situation from the fifties on. In particular papers by Panov, by = Mel'chuk by Zhirkov and by Bel'skaja. I am having a bit of trouble = finding these and other authors writings. Can anyone point out where I = can find electronic versions perhaps? Thank you Mike Corradi ------=_NextPart_000_0011_01C34F66.E6BCD7A0 Content-Type: text/html; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable
Dear All,
 
I am currently working on my thesis at the = University=20 of Parma, it is on MT with particular reference to Italian and Russian. = At the=20 moment I am working on a short history section and I am looking for = material on=20 the Russian situation from the fifties on. In particular papers by = Panov, by=20 Mel'chuk  by Zhirkov and by Bel'skaja. I am having a bit of trouble = finding=20 these and other authors writings. Can anyone point out where I can find=20 electronic versions perhaps?
 
Thank you
 
Mike Corradi
------=_NextPart_000_0011_01C34F66.E6BCD7A0-- From WJHutchins@compuserve.com Mon Jul 21 18:11:56 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Mon, 21 Jul 2003 13:11:56 -0400 Subject: [MT-List] Russian researchers Message-ID: <200307211312_MC3-1-4387-980F@compuserve.com> Dear Mike Corradi, I think it is very unlikely that you will find electronic versions of these papers. I have photocopies of a number of Russian papers from the period. Let me know which ones you want and I can photocopy them for you. Best wishes, John Hutchins PS. You might find my 1986 book on the history of MT of some use to you. It is now all available (as PDF files) on my website at http://ourworld.compuserve.com/homepages/WJHutchins =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D >Dear All, = I am currently working on my thesis at the University of Parma, it is on = MT with particular reference to Italian and Russian. At the moment I am working on a short history section and I am looking for material on the Russian situation from the fifties on. In particular papers by Panov, by Mel'chuk by Zhirkov and by Bel'skaja. I am having a bit of trouble findi= ng these and other authors writings. Can anyone point out where I can find electronic versions perhaps? Thank you Mike Corradi< From abcs@online.no Tue Jul 22 22:00:20 2003 From: abcs@online.no (=?iso-8859-1?Q?=D8yvind_Str=F8m?=) Date: Tue, 22 Jul 2003 23:00:20 +0200 Subject: SV: [MT-List] Russian researchers In-Reply-To: <001401c34f56$23d2b8a0$fa4fb450@mike> Message-ID: <000001c35094$4446d320$ae604382@OSVAIO2> This is a multi-part message in MIME format. ------=_NextPart_000_0001_01C350A5.07CFA320 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hello Mike =20 It might be of interest to contact Michael Blekhman of PARS, a company with several years of history in russian MT. ling98@canada.com =20 He seem to be very serviceminded. =20 Kind regards =20 =D8yvind Str=F8m =20 =20 =20 Dear All,=20 =20 I am currently working on my thesis at the University of Parma, it is on MT with particular reference to Italian and Russian. At the moment I am working on a short history section and I am looking for material on the Russian situation from the fifties on. In particular papers by Panov, by Mel'chuk by Zhirkov and by Bel'skaja. I am having a bit of trouble finding these and other authors writings. Can anyone point out where I can find electronic versions perhaps? =20 Thank you =20 Mike Corradi ------=_NextPart_000_0001_01C350A5.07CFA320 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

Hello = Mike

 

It might be of interest to = contact

Michael Blekhman

of PARS, a company with several years of history in russian = MT.

ling98@canada.com=

 

He seem to be very = serviceminded.

 

Kind = regards

 

=D8yvind = Str=F8m

 

 

 

=

Dear All, =

 

I = am currently working on my thesis at the University of Parma, it is on MT with = particular reference to Italian and Russian. At the moment I am working on a short = history section and I am looking for material on the Russian situation from the = fifties on. In particular papers by Panov, by Mel'chuk  by Zhirkov and by Bel'skaja. I am having a bit of trouble finding these and other authors writings. Can anyone point out where I can find electronic versions = perhaps?

 

Thank = you

 

Mike = Corradi

------=_NextPart_000_0001_01C350A5.07CFA320-- From cb@lim.nl Fri Aug 1 10:01:57 2003 From: cb@lim.nl (Colin Brace) Date: Fri, 01 Aug 2003 11:01:57 +0200 Subject: [MT-List] IHT: "Big leap in machine language translation" Message-ID: Hi all, There is an article in today's International Herald Tribune on machine translation (maybe it is in the NYT as well): "Statistical machine translation - in which computers essentially learn new languages on their own instead of being "taught" the languages by bilingual human programmers - has taken off." ... The rest: http://www.iht.com/articles/104853.html Enjoy! -- Colin Brace | cb@lim.nl Amsterdam http://www.lim.nl From dekai@cs.ust.hk Fri Aug 1 11:00:55 2003 From: dekai@cs.ust.hk (dekai@cs.ust.hk) Date: Fri, 1 Aug 2003 06:00:55 -0400 (EDT) Subject: [MT-List] NYTimes.com Article: From Uzbek to Klingon, the Machine Cracks the Code Message-ID: <20030801100055.B059F35042@web38t.prvt.nytimes.com> This article from NYTimes.com has been sent to you by dekai@cs.ust.hk. The NY Times version, entitled "From Uzbek to Klingon, the Machine Cracks the Code", was slightly longer. http://www.nytimes.com/2003/07/31/technology/circuits/31next.html -Dekai Dr. Dekai Wu, Human Language Technology Center, HKUST Dept of Computer Science, Univ of Science & Technology, Clear Water Bay, Hong Kong tel +852 2358.7000 / dir +852 2358.6989 / fax +852 2358.1477 dekai@cs.ust.hk / www.cs.ust.hk/~dekai ----- Original Message ----- From: "Colin Brace" To: Sent: Friday, August 01, 2003 5:01 PM Subject: [MT-List] IHT: "Big leap in machine language translation" > Hi all, > > There is an article in today's International Herald Tribune on machine > translation (maybe it is in the NYT as well): > > "Statistical machine translation - in which computers essentially learn > new languages on their own instead of being "taught" the languages by > bilingual human programmers - has taken off." ... > > The rest: > > http://www.iht.com/articles/104853.html > > Enjoy! dekai@cs.ust.hk /-------------------- advertisement -----------------------\ Explore more of Starbucks at Starbucks.com. http://www.starbucks.com/default.asp?ci=1015 \----------------------------------------------------------/ >From Uzbek to Klingon, the Machine Cracks the Code July 31, 2003 By CHRISTOPHER JOHN FARAH IN the summer of 1999, at a workshop on statistical machine translation at Johns Hopkins University, Kevin Knight passed out a copy of an advertisement to each member of the research team he was leading. In the center of the ad was a picture of a yellowed, frayed parchment covered in Japanese characters. "To most people, this looks like a secret code," the ad announced. "Codes are meant to be broken." The ad was for a product yet to be created called the Decoder. "Pour in a new bunch of text," said the ad's text, alongside a picture of a software box. "We think you'll be surprised." The Decoder was meant to be a motivational tool. At the time, the field of statistical machine translation was all but dead. In the four years that have passed since that workshop, Dr. Knight, the head of machine translation research at the University of Southern California's Information Sciences Institute, is amazed by just how prophetic the ad has proved. "Here we are," he said. "It's no joke anymore." Statistical machine translation - in which computers essentially learn new languages on their own instead of being "taught" the languages by bilingual human programmers - has taken off. The new technology allows scientists to develop machine translation systems for a wide number of obscure languages at a pace that experts once thought impossible. Dr. Knight and others said the progress and accuracy of statistical machine translation had recently surpassed that of the traditional machine translation programs used by Web sites like Yahoo and BabelFish. In the past, such programs were able to compile extensive databanks of foreign languages that allowed them to outperform statistics-based systems. Traditional machine translation relies on painstaking efforts by bilingual programmers to enter the vast wealth of information on vocabulary and syntax that the computer needs to translate one language into another. But in the early 1990's, a team of researchers at I.B.M. devised another way to do things: feeding a computer an English text and its translation in a different language. The computer then uses statistical analysis to "learn" the second language. Compare two simple phrases in Arabic: "rajl kabir'' and "rajl tawil.'' If a computer knows that the first phrase means "big man," and the second means "tall man," the machine can compare the two and deduce that rajl means "man," while kabir and tawil mean "big" and "tall," respectively. Phrases like these, called "N-grams" (with N representing the number of terms in a given phrase) are the basic building blocks of statistical machine translation. Although in one sense it was more economical, this kind of machine translation was also much more complex, requiring powerful computers and software that did not exist for most of the 90's. The Johns Hopkins workshop changed all that, yielding a software application package, Egypt/Giza, that made statistical translation accessible to researchers across the country. "We wanted to jump-start a vibrant field," Dr. Knight said. "There was no software or data to play with." Today researchers are racing to improve the quality and accuracy of the translations. The final translations generally give an average reader a solid understanding of the original meaning but are far from grammatically correct. While not perfect, statistics-based technology is also allowing scientists to crack scores of languages in a fraction of the time, and at a fraction of the cost, that traditional methods involved. A team of computer scientists at Johns Hopkins led by David Yarowsky is developing machine translations of such languages as Uzbek, Bengali, Nepali - and one from "Star Trek." "If we can learn how to translate even Klingon into English, then most human languages are easy by comparison," he said. "All our techniques require is having texts in two languages. For example, the Klingon Language Institute translated 'Hamlet' and the Bible into Klingon, and our programs can automatically learn a basic Klingon-English MT system from that.'' Dr. Yarowsky said he hoped to have working translation systems for as many as 100 languages within five years. Although the grammatical structures of languages like Chinese and Arabic make them hard to analyze statistically, he said, it will only be a matter of time before such hurdles are overcome. "At some point, we start encountering the same problems over and over," he said. In addition to the release of Egypt/Giza in 1999, the spread of the Internet has led to an explosion of translated texts in far-flung languages, greatly aiding the team's research. Researchers have also benefited from a much faster means of evaluating the outcome of translation experiments: a computerized technique developed by I.B.M. enables researchers to test 10 to 100 new approaches for cracking languages each day. The technique, known as the Bleu Metric, compares machine translations with a "gold standard" based on human translations. Instead of waiting for human beings to assign a score to the quality of a machine translation, the Bleu Metric does so almost instantly through a statistical comparison. This provides scientists with a fast, objective measurement that they can use to note improvement and saves them from having to review every unsuccessful experiment. "Before Bleu, it was really a bad state of affairs," said Alex Fraser, a doctoral student at U.S.C. "You look at broken couplets of English for a long time, and eventually you start to accept it more and more." Despite the progress being made in statistical machine translation, some researchers remain skeptical, preferring to focus their efforts on language-specific translation techniques. Ophir Frieder, a professor of computer science at the Illinois Institute of Technology, is working on a search system exclusive to Arabic text. "Yes, N-grams work on any language, but as a search technique they work poorly on every language," he said. "It's a basic novice solution." Dr. Knight acknowledges that statistical machine translation is far from perfect. In its latest efforts, his team has sought to combine the statistical and traditional approaches to achieve maximum accuracy and to produce translations that the average computer user can understand. The best machine translation systems today, while capable of yielding a passage's general meaning, are better known for their muddled syntax than their accuracy. By applying the principles of statistical translation to varying grammatical structures, Dr. Knight hopes to resolve some of these basic problems. "N-grams are one of those things where you don't know how much you need it until you take it away," he said. "The way our imaginations work, we need help." http://www.nytimes.com/2003/07/31/technology/circuits/31next.html?ex=1060732055&ei=1&en=72d31e0bcb2bad2d --------------------------------- Get Home Delivery of The New York Times Newspaper. Imagine reading The New York Times any time & anywhere you like! Leisurely catch up on events & expand your horizons. Enjoy now for 50% off Home Delivery! Click here: http://www.nytimes.com/ads/nytcirc/index.html HOW TO ADVERTISE --------------------------------- For information on advertising in e-mail newsletters or other creative advertising opportunities with The New York Times on the Web, please contact onlinesales@nytimes.com or visit our online media kit at http://www.nytimes.com/adinfo For general information about NYTimes.com, write to help@nytimes.com. Copyright 2003 The New York Times Company From WJHutchins@compuserve.com Mon Aug 4 12:20:48 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Mon, 4 Aug 2003 07:20:48 -0400 Subject: [MT-List] update of my website Message-ID: <200308040721_MC3-1-45BB-D4FA@compuserve.com> Those who are interesed in the history of machine translation may like to know that all chapters of my 1986 book are now available (as PDF files) on my website: http://ourworld.compuserve.com/homepages/WJHutchins You will also find other recent additions of newer and older papers. John Hutchins 4 August From lorena.guerra2@mail.dcu.ie Wed Aug 6 11:32:14 2003 From: lorena.guerra2@mail.dcu.ie (lorena.guerra2@mail.dcu.ie) Date: Wed, 6 Aug 2003 11:32:14 +0100 Subject: [MT-List] Pricing Post-editing of MT Message-ID: <3EFBE21800020E12@hawk.dcu.ie> Dear all, I'm looking for information about the prices of post-editing of raw MT ou= tput. Whereas human translation is mainly based on the unit "word" as a cost ba= se, in the case of the post-editor, which is the situation? how much is s/he paid, more or less? Thanks, Lorena Postgraduate student at DCU- Ireland Information needed for my dissertation in Translation Technology From henrik.wilhelm.gade@medtronic.com Wed Aug 6 12:17:18 2003 From: henrik.wilhelm.gade@medtronic.com (Henrik Gade) Date: Wed, 06 Aug 2003 06:17:18 -0500 Subject: [MT-List] Pricing Post-editing of MT Message-ID: Dear Lorena, We normally pay per hour for post-editing of raw MT-output, the list price = being approx. 70 euro per hour subject to the agencies we use. Best regards Henrik W. Gade The R&D Department, Medtronic A/S, Copenhagen >>> 08/06/03 12:32pm >>> Dear all, I'm looking for information about the prices of post-editing of raw MT = output. Whereas human translation is mainly based on the unit "word" as a cost = base, in the case of the post-editor, which is the situation? how much is s/he paid, more or less? Thanks, Lorena Postgraduate student at DCU- Ireland Information needed for my dissertation in Translation Technology -- For MT-List info, see http://www.eamt.org/mt-list.html From daniel.grasmick@sap.com Wed Aug 6 13:21:14 2003 From: daniel.grasmick@sap.com (Grasmick, Daniel) Date: Wed, 6 Aug 2003 14:21:14 +0200 Subject: [MT-List] Pricing Post-editing of MT Message-ID: <22772E6015DA354B84C6F8DB8BF60504093344BC@dewdfx13.wdf.sap.corp> Hi Lorena and Henrik, In our case we combine TM and MT for most of the translation projects - and deliver translation memories containing also raw MT segments to our agencies for post-editing. MT matches are treated like TM fuzzy matches of category 2 and paid about* 60% of the normal price. Hendrik, how do you reward the translators doing post-editing of TM fuzzy matches? Best, Daniel * about 60% since this has to be negociated and depends a) on the quality of the source and b) obviously also on the quality and maturity of the MT system used. -----Original Message----- From: Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com] Sent: Mittwoch, 6. August 2003 13:17 To: mt-list@eamt.org; lorena.guerra2@mail.dcu.ie Subject: Re: [MT-List] Pricing Post-editing of MT Dear Lorena, We normally pay per hour for post-editing of raw MT-output, the list price being approx. 70 euro per hour subject to the agencies we use. Best regards Henrik W. Gade The R&D Department, Medtronic A/S, Copenhagen >>> 08/06/03 12:32pm >>> Dear all, I'm looking for information about the prices of post-editing of raw MT output. Whereas human translation is mainly based on the unit "word" as a cost base, in the case of the post-editor, which is the situation? how much is s/he paid, more or less? Thanks, Lorena Postgraduate student at DCU- Ireland Information needed for my dissertation in Translation Technology -- For MT-List info, see http://www.eamt.org/mt-list.html -- For MT-List info, see http://www.eamt.org/mt-list.html From henrik.wilhelm.gade@medtronic.com Wed Aug 6 13:58:44 2003 From: henrik.wilhelm.gade@medtronic.com (Henrik Gade) Date: Wed, 06 Aug 2003 07:58:44 -0500 Subject: [MT-List] Pricing Post-editing of MT Message-ID: Dear Daniel, We regard post-editing as a qualified human translation work with a little = MT help to speed up the process, We are producing literally hundreds of = manuals a year, so speed is essential. We don't use fuzzy matches - = Systranet produces really good raw texts, if we fine tune the English = original. Best regards Henrik=20 >>> "Grasmick, Daniel" 08/06/03 02:21pm >>> Hi Lorena and Henrik, In our case we combine TM and MT for most of the translation projects - = and deliver translation memories containing also raw MT segments to our = agencies for post-editing. MT matches are treated like TM fuzzy matches of category 2 and paid about* = 60% of the normal price. Hendrik, how do you reward the translators doing post-editing of TM fuzzy = matches? Best, Daniel * about 60% since this has to be negociated and depends a) on the quality = of the source and b) obviously also on the quality and maturity of the MT = system used. -----Original Message----- From: Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com]=20 Sent: Mittwoch, 6. August 2003 13:17 To: mt-list@eamt.org; lorena.guerra2@mail.dcu.ie=20 Subject: Re: [MT-List] Pricing Post-editing of MT Dear Lorena, We normally pay per hour for post-editing of raw MT-output, the list price = being approx. 70 euro per hour subject to the agencies we use. Best regards Henrik W. Gade The R&D Department, Medtronic A/S, Copenhagen >>> 08/06/03 12:32pm >>> Dear all, I'm looking for information about the prices of post-editing of raw MT = output. Whereas human translation is mainly based on the unit "word" as a cost = base, in the case of the post-editor, which is the situation? how much is s/he paid, more or less? Thanks, Lorena Postgraduate student at DCU- Ireland Information needed for my dissertation in Translation Technology -- For MT-List info, see http://www.eamt.org/mt-list.html=20 --=20 For MT-List info, see http://www.eamt.org/mt-list.html From daniel.grasmick@sap.com Wed Aug 6 16:56:47 2003 From: daniel.grasmick@sap.com (Grasmick, Daniel) Date: Wed, 6 Aug 2003 17:56:47 +0200 Subject: [MT-List] Pricing Post-editing of MT Message-ID: <22772E6015DA354B84C6F8DB8BF60504093344C9@dewdfx13.wdf.sap.corp> Hi Henrik, Thank you for this lively exchange. :-) We have started to combine MT and TM for the translation of mass data in 1993-94 and have certainly translated and edited (internally and externally) well other than 100 million segments. Our philosophy is the following: - a segment edited by a human translator is generally of better value than an MT raw segment. This is why we extract only the source segments having a match value under 75% - compared to the specific memory for this project. - after having machine-translated the delta, we reimport the MT raw segments (that have normally only been "machine edited" using pattern matching rules). The MT segments get a penalty of 15% - and are therefore always listed before fuzzy matches of values under 75%. - we think that a good MT match is very often better that an average fuzzy match but prefer to leave the decision to the post-editors. In our opinion, both technologies have always been complementary and do not exclude each other. As far as I know, Medtronics is also a big user of TM. Do you know Peter Wilms van Kersbergen? Maybe we should exchange other ideas "offline" and not hijack Lorena's original mail... (Sorry if other people felt bored...) Thank you again and best regards, Daniel -----Original Message----- From: Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com] Sent: Mittwoch, 6. August 2003 14:59 To: lorena.guerra2@mail.dcu.ie; Grasmick, Daniel Cc: mt-list@eamt.org Subject: RE: [MT-List] Pricing Post-editing of MT Dear Daniel, We regard post-editing as a qualified human translation work with a little MT help to speed up the process, We are producing literally hundreds of manuals a year, so speed is essential. We don't use fuzzy matches - Systranet produces really good raw texts, if we fine tune the English original. Best regards Henrik >>> "Grasmick, Daniel" 08/06/03 02:21pm >>> Hi Lorena and Henrik, In our case we combine TM and MT for most of the translation projects - and deliver translation memories containing also raw MT segments to our agencies for post-editing. MT matches are treated like TM fuzzy matches of category 2 and paid about* 60% of the normal price. Hendrik, how do you reward the translators doing post-editing of TM fuzzy matches? Best, Daniel * about 60% since this has to be negociated and depends a) on the quality of the source and b) obviously also on the quality and maturity of the MT system used. -----Original Message----- From: Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com] Sent: Mittwoch, 6. August 2003 13:17 To: mt-list@eamt.org; lorena.guerra2@mail.dcu.ie Subject: Re: [MT-List] Pricing Post-editing of MT Dear Lorena, We normally pay per hour for post-editing of raw MT-output, the list price being approx. 70 euro per hour subject to the agencies we use. Best regards Henrik W. Gade The R&D Department, Medtronic A/S, Copenhagen >>> 08/06/03 12:32pm >>> Dear all, I'm looking for information about the prices of post-editing of raw MT output. Whereas human translation is mainly based on the unit "word" as a cost base, in the case of the post-editor, which is the situation? how much is s/he paid, more or less? Thanks, Lorena Postgraduate student at DCU- Ireland Information needed for my dissertation in Translation Technology -- For MT-List info, see http://www.eamt.org/mt-list.html -- For MT-List info, see http://www.eamt.org/mt-list.html From esteam@otenet.gr Wed Aug 6 16:58:41 2003 From: esteam@otenet.gr (ESTeam) Date: Wed, 06 Aug 2003 18:58:41 +0300 Subject: [MT-List] Pricing Post-editing of MT References: <22772E6015DA354B84C6F8DB8BF60504093344C9@dewdfx13.wdf.sap.corp> Message-ID: <3F3125B1.37EDFD88@otenet.gr> Dear Daniel and Henrik, as a developer of integrated TM and MT I couldn't agree more with Daniel. I hope you, Henrik, do take his advice and use the postedited MT as a TM resource. Thank you for the interesting dialogue, Gudrun Magnusdottir ESTeam AB "Grasmick, Daniel" wrote: > Hi Henrik, > > Thank you for this lively exchange. > :-) > We have started to combine MT and TM for the translation of mass data in 1993-94 and have certainly translated and edited (internally and externally) well other than 100 million segments. > > Our philosophy is the following: > - a segment edited by a human translator is generally of better value than an MT raw segment. This is why we extract only the source segments having a match value under 75% - compared to the specific memory for this project. > - after having machine-translated the delta, we reimport the MT raw segments (that have normally only been "machine edited" using pattern matching rules). The MT segments get a penalty of 15% - and are therefore always listed before fuzzy matches of values under 75%. > - we think that a good MT match is very often better that an average fuzzy match but prefer to leave the decision to the post-editors. > In our opinion, both technologies have always been complementary and do not exclude each other. > As far as I know, Medtronics is also a big user of TM. > Do you know Peter Wilms van Kersbergen? > Maybe we should exchange other ideas "offline" and not hijack Lorena's original mail... > (Sorry if other people felt bored...) > Thank you again and best regards, > Daniel > > -----Original Message----- > From: Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com] > Sent: Mittwoch, 6. August 2003 14:59 > To: lorena.guerra2@mail.dcu.ie; Grasmick, Daniel > Cc: mt-list@eamt.org > Subject: RE: [MT-List] Pricing Post-editing of MT > > Dear Daniel, > > We regard post-editing as a qualified human translation work with a little MT help to speed up the process, We are producing literally hundreds of manuals a year, so speed is essential. We don't use fuzzy matches - Systranet produces really good raw texts, if we fine tune the English original. > > Best regards > Henrik > > >>> "Grasmick, Daniel" 08/06/03 02:21pm >>> > Hi Lorena and Henrik, > > In our case we combine TM and MT for most of the translation projects - and deliver translation memories containing also raw MT segments to our agencies for post-editing. > MT matches are treated like TM fuzzy matches of category 2 and paid about* 60% of the normal price. > > Hendrik, how do you reward the translators doing post-editing of TM fuzzy matches? > > Best, > Daniel > * about 60% since this has to be negociated and depends a) on the quality of the source and b) obviously also on the quality and maturity of the MT system used. > > -----Original Message----- > From: Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com] > Sent: Mittwoch, 6. August 2003 13:17 > To: mt-list@eamt.org; lorena.guerra2@mail.dcu.ie > Subject: Re: [MT-List] Pricing Post-editing of MT > > Dear Lorena, > > We normally pay per hour for post-editing of raw MT-output, the list price being approx. 70 euro per hour subject to the agencies we use. > > Best regards > Henrik W. Gade > > The R&D Department, Medtronic A/S, Copenhagen > > >>> 08/06/03 12:32pm >>> > Dear all, > > I'm looking for information about the prices of post-editing of raw MT output. > Whereas human translation is mainly based on the unit "word" as a cost base, > in the case of the post-editor, which is the situation? how much is s/he > paid, more or less? > Thanks, > > Lorena > Postgraduate student at DCU- Ireland > Information needed for my dissertation in Translation Technology > > -- > For MT-List info, see http://www.eamt.org/mt-list.html > > -- > For MT-List info, see http://www.eamt.org/mt-list.html > > -- > For MT-List info, see http://www.eamt.org/mt-list.html From hjin@ic.sunysb.edu Thu Aug 7 20:55:34 2003 From: hjin@ic.sunysb.edu (Huiling Jin) Date: Thu, 7 Aug 2003 15:55:34 -0400 (EDT) Subject: [MT-List] help with giza++ Message-ID: Dear All, I am working on a MT project, which uses Giza++ to build the translation model, and I have some questions about it: If I make changes in the alignment file (*.A3.*), which is one of Giza++'s output files, would it affect the decoding subsequently? If not, is there any way (or any output files that I can change) to improve the alignment in the translation model? Also, when Giza++ talks about source language and target language, does it mean translation source and target, or NCM (Noise Channel Model) source and target, which is the other way around. For example, if I am translating from English to Chinese, should it be source:English, target:Chinese, as in the normal translation terminology; or source:Chinese, target:English, as in the NCM terminology? Thanks in advance. Huiling Graduate Student Computer Science Dept. SUNY at Stony Brook USA From hjin@ic.sunysb.edu Mon Aug 11 18:03:50 2003 From: hjin@ic.sunysb.edu (Huiling Jin) Date: Mon, 11 Aug 2003 13:03:50 -0400 (EDT) Subject: [MT-List] (no subject) Message-ID: Hi, All, Many thanks for your replies and for sharing the information. It's been very helpful :) . Paul, So far, I only have about 70,000 word data to train the Chinese language model. I also have the problem of lack of bitext since I am doing domain-specific SMT. Huiling From hjin@ic.sunysb.edu Wed Aug 13 05:20:18 2003 From: hjin@ic.sunysb.edu (Huiling Jin) Date: Wed, 13 Aug 2003 00:20:18 -0400 (EDT) Subject: [MT-List] summary: questions on giza++ Message-ID: Dear All, Thanks to everyone who responded to my mail on questions on giza++. Here is a summary of the correspondences: Question 1: > I am working on a MT project, which uses Giza++ to build the translation > model, and I have some questions about it: If I make changes in the > alignment file (*.A3.*), which is one of Giza++'s output files, would it > affect the decoding subsequently? If not, is there any way (or any > output files that I can change) to improve the alignment in the > translation model? No, *.A3.* file include only sample of the best (Vitebi) alignment for each sentence, it doesn't have any influence on decoding process. Decoding with Model3 uses: *.a3.final, *.d3.final, *.n3.final, *.p0_3.final, *.t3.final Decoding with Model4 uses: *.a3.final, *.d4.final, *.n3.final, *.p0_3.final, *.t3.final Thanks to Jan Curin --------------------------- Alignment can be improved by the use of a dictionary file which biases the alignment model. Thanks to Paul Johnston ---------------------------- One could tweak the probability files, which is used by the ISI decoder. But it's probably better to improve alignments by adjusting the number of training iterations one uses for each of the IBM Models, or by doing pre- and post-processing of the training data. Thanks to Chris Callison-Burch -------------------------------------- -------------------------------------- Question 2: > Also, when Giza++ talks about source language and target language, does it > mean translation source and target, or NCM (Noise Channel > Model) source and target, which is the other way around. For example, if I > am translating from English to Chinese, should it be source:English, > target:Chinese, as in the normal translation terminology; or > source:Chinese, target:English, as in the NCM terminology? Giza++ uses the NCM definition of source and target, therefore, in the above example, Chinese is the source and English the target. Thanks to Jan Curin, Paul Johnston and Chris Callison-Burch ----------------- Also thank Colin Brace for his kind reminder :) ************************* Huiling Jin CS Dept. SUNY Stony Brook From WJHutchins@compuserve.com Thu Aug 28 10:59:23 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Thu, 28 Aug 2003 05:59:23 -0400 Subject: [MT-List] Compendium of translation software: NEW edition Message-ID: <200308280559_MC3-1-4ABB-129F@compuserve.com> The seventh edition (August 2003) of the "Compendium of translation software" is now available from the EAMT website (www.eamt.org). Existing users can continue to = use their old log on procedures. (New users will need to = obtain password etc. from the EAMT secretariat.) As reported before, older (superseded) editions of the Compendium can be consulted from my own website: http://ourworld.compuserve.com/homepages/WJHutchins Please send me information about new products, any changes, notices of products no longer marketed, etc. The next edition is being planned for December or January. John Hutchins 28 Aug = From jmont@tin.it Thu Sep 25 08:46:24 2003 From: jmont@tin.it (Johanna) Date: Thu, 25 Sep 2003 09:46:24 +0200 Subject: [MT-List] MT on-line services Message-ID: <000801c38339$22cf6ea0$7bbd6850@johmonti> This is a multi-part message in MIME format. ------=_NextPart_000_0005_01C38349.E230B500 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable My name is Johanna Monti and I teach History and Theory of Translation = at the University L'Orientale of Neaples.=20 I'm looking for the following infomation: how many users use MT on-line services ? how many words/lines, etc. are translated per day/year by MT on-line = services? what type of texts are translated and in which percentage? what is the purpose of the translation (internal use, for information = dissemination, etc.)? Does anybody know if there are any articles/figures available about this = issue? Thanks for your help. Johanna Monti ------=_NextPart_000_0005_01C38349.E230B500 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
My name is Johanna Monti and I teach = History and=20 Theory of Translation at the University L'Orientale of Neaples. =
I'm looking for the following=20 infomation:
how many users use MT on-line services=20 ?
how many words/lines, etc. are = translated per=20 day/year by MT on-line services?
what type of texts are translated and = in which=20 percentage?
what is the purpose of the translation = (internal=20 use, for information dissemination, etc.)?
 
Does anybody know if there are any = articles/figures=20 available about this issue?
 
Thanks for your help.
 
Johanna = Monti
------=_NextPart_000_0005_01C38349.E230B500-- From jeff.allen@free.fr Sat Sep 27 16:18:05 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Sat, 27 Sep 2003 17:18:05 +0200 Subject: [MT-List] more MT software review articles References: <00da01c2f8df$d1e47140$0100007f@home> <07a101c3081b$bbf76e00$df9c933e@home> Message-ID: <016c01c3850b$42d510a0$1699933e@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_0140_01C3851B.504F6EA0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear all, =20 A few software review articles have appeared lately:=20 GUERRA, Lorena. 2003. Review: Machine Translation with @promt = Professional. Irish Translators' & Interpreters' Association (ITIA) = Bulletin. September 2003. pp. 5-6. WASSMER, Thomas. 2003. Review: Systran 4.0: Personal, Standard, = Premium. In Multilingual Computing and Technology magazine. Number 58, = Vol. 14, Issue 6. September 2003. Pp. 17-19. This should be available online at: http://www.multilingual.com/ ALLEN, Jeff. 2003. Review: Systran Premium 4.0. In Multilingual = Computing and Technology magazine. Number 58, Vol. 14, Issue 6. = September 2003. Pp. 19-22. This should be available online at: http://www.multilingual.com/ Also, a thesis/dissertation: GUERRA, Lorena. 2003. Human Translation versus Machine Translation and = Full Post-Editing of Raw Machine Translation Output. Master's = Dissertation. Dublin City University. September 2003.=20 Regards, Jeff --------------------------------- Jeff ALLEN Paris, France Mobile Tel: (+33) 6 86 06 87 59 e-mail: jeff.allen@free.fr =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Saturday, April 19, 2003 5:02 PM Objet : [MT-List] another MT software review article ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Wednesday, April 02, 2003 7:47 AM Objet : [MT-List] more articles in MLCT on MT and translation = technologies =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Friday, March 14, 2003 10:40 PM Objet : [MT-List] updated links: articles in MLCT on MT, TM, = postediting ------=_NextPart_000_0140_01C3851B.504F6EA0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Dear all,
 
A few software review articles have appeared = lately:=20
 
GUERRA, Lorena. 2003.  Review: Machine Translation with @promt = Professional.  Irish Translators' & Interpreters' Association = (ITIA)=20 Bulletin. September 2003. pp. 5-6.
 
WASSMER, Thomas.  2003.  Review:  = Systran 4.0:=20 Personal, Standard, Premium.  In Multilingual Computing and = Technology=20 magazine. Number 58, Vol. 14, Issue 6. September 2003. Pp. = 17-19.
This should be available online at: http://www.multilingual.com/
 
ALLEN, Jeff. 2003. Review:  Systran Premium = 4.0.  In=20 Multilingual Computing and Technology magazine. Number 58, Vol. 14, = Issue 6.=20 September 2003. Pp. 19-22.
This should be available online at: http://www.multilingual.com/
 
Also, a thesis/dissertation:
 
GUERRA, Lorena. 2003.  Human Translation versus Machine = Translation=20 and Full Post-Editing of Raw Machine Translation Output.  Master's=20 Dissertation. Dublin City University.  September 2003.
 
 
Regards,
Jeff

---------------------------------
Jeff = ALLEN
Paris,=20 France
Mobile Tel: (+33) 6 86 06 87 59
e-mail: jeff.allen@free.fr
 
----- Message d'origine -----
De=20 : Jeff=20 Allen
Envoy=E9 : Saturday, April = 19, 2003=20 5:02 PM
Objet : [MT-List] another MT = software=20 review article
 
<snip>
 
----- Message d'origine -----
De=20 : Jeff=20 Allen
Envoy=E9 : Wednesday, = April 02, 2003=20 7:47 AM
Objet : [MT-List] more = articles in MLCT=20 on MT and translation technologies

<snip>
 
  ----- = Message=20 d'origine -----
De : Jeff = Allen=20
Envoy=E9 : Friday, March 14, 2003 10:40 PM
Objet : [MT-List] updated links: articles in MLCT on MT, = TM,=20 postediting

<snip>
------=_NextPart_000_0140_01C3851B.504F6EA0-- From jeff.allen@free.fr Sun Sep 28 15:29:27 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Sun, 28 Sep 2003 16:29:27 +0200 Subject: [MT-List] MT on-line services References: <000801c38339$22cf6ea0$7bbd6850@johmonti> Message-ID: <037d01c385cd$3e1d3fc0$449a933e@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_0362_01C385DD.AF9BDD40 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Johanna Monti wrote: >how many words/lines, etc. are translated per day/year by MT on-line = services? Johanna, Some published statement below concerning online MT systems. =20 ----- * Systran: According to MT specialist Scott Bennett, "Altavista's = BabelFish... initiated in late 1997, now is used a million times per = day, according to the Systran site." (Bennett, Scott. 2000. "Taking the = Babble out of Babel Fish". Industry pioneer Scott Bennett reviews = Systran's Machine Translation implementation on AltaVista. In special = issue on Machine Translation in Language International magazine, 2000. = Vol. 12, No. 3, pp. 20-21) * Softissimo, MT value added reseller company, announced at a public = demonstration on 12 July 2000 at the Hotel Bristol in Paris, France that = their Reverso MT portal receives over ten thousand (10,000) free = translation requests per day. * Softissimo adds to Reverso Pro/Mac; increases Internet translation = volume; section: Product Watch & Industry Outlook. In Multilingual = Computing and Technology magazine. Number 40, Vol. 12, Issue 4. June = 2001. p. 19. "Softissimo also announced that the Internet translation request volume = processed by its Reverso translation engine (www.reverso.net) has now = reached several million translation requests of Web pages, e-mail, short = texts and results of search engine requests) per month on its mail = translation portal and the portals of its Internet partners." * Jaap van der Meer. 2003. The Business Case for MT The Breakthrough = Is for Real. In LISA Newsletter XII-2_6.=20 "Every day, portals like Altavista and Google process nearly ten million = requests for automatic translation." * The Machine Translation Workshop of TALN'2001=20 Subject: [MT-List] DEADLINE EXTENSION - Workshop on Machine Translation Date: Sat, 14 Apr 2001 09:21:00 -0700 From: Christiane PANISSOD "Indeed, MT has seen a remarkable development for the past few years in = terms of the number of translation requests (about a million translation = requests everyday on the Web) as well as in terms of the different types = of formats translated : translation of dynamic resources (FAQ, Web = pages, daily papers), emails, requests for search engines, etc. " ------------ Regards, Jeff Allen ----- Message d'origine -----=20 De : Johanna=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Thursday, September 25, 2003 9:46 AM Objet : [MT-List] MT on-line services My name is Johanna Monti and I teach History and Theory of Translation = at the University L'Orientale of Neaples.=20 I'm looking for the following infomation: how many users use MT on-line services ? how many words/lines, etc. are translated per day/year by MT on-line = services? what type of texts are translated and in which percentage? what is the purpose of the translation (internal use, for information = dissemination, etc.)? =20 Does anybody know if there are any articles/figures available about = this issue? =20 Thanks for your help. =20 Johanna Monti ------=_NextPart_000_0362_01C385DD.AF9BDD40 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Johanna Monti wrote:
>how many words/lines, = etc. are=20 translated per day/year by MT on-line services?
 
Johanna,
 
Some published statement below = concerning online MT=20 systems.
 
-----
* Systran: According to MT specialist Scott Bennett, = "Altavista's BabelFish... initiated in late 1997, now is used a million = times=20 per day, according to the Systran site." (Bennett, Scott. 2000. "Taking = the=20 Babble out of Babel Fish". Industry pioneer Scott Bennett reviews = Systran's=20 Machine Translation implementation on AltaVista. In special issue on = Machine=20 Translation in Language International magazine, 2000. Vol. 12, No. 3, = pp.=20 20-21)
 
* Softissimo, MT value added reseller company, = announced at a=20 public demonstration on 12 July 2000 at the Hotel Bristol in Paris, = France that=20 their Reverso MT portal receives over ten thousand (10,000) free = translation=20 requests per day.
 
*  Softissimo adds to Reverso Pro/Mac; = increases Internet=20 translation volume;  section:  Product Watch & Industry=20 Outlook.  In Multilingual Computing and Technology magazine. Number = 40,=20 Vol. 12, Issue 4. June 2001. p. 19.
"Softissimo also announced that the Internet = translation=20 request volume processed by its Reverso translation engine (www.reverso.net) has now reached = several=20 million translation requests of Web pages, e-mail, short texts and = results of=20 search engine requests) per month on its mail translation portal and the = portals=20 of its Internet partners."
 
*  Jaap van der Meer.  2003.  The = Business Case=20 for MT The Breakthrough Is for Real. In LISA Newsletter XII-2_6. =
"Every day, portals like Altavista and Google = process nearly=20 ten million requests for automatic translation."
 
*  The Machine Translation Workshop of = TALN'2001=20
 
Subject: [MT-List] DEADLINE EXTENSION - Workshop on Machine=20 Translation
Date: Sat, 14 Apr 2001 09:21:00 -0700
From: Christiane = PANISSOD
 
"Indeed, MT has seen a remarkable development for = the past few=20 years in terms of the number of translation requests (about a million=20 translation requests everyday on the Web) as well as in terms of the = different=20 types of formats translated : translation of dynamic resources (FAQ, Web = pages,=20 daily papers), emails, requests for search engines, etc. "
------------
 
Regards,
 
Jeff Allen
 
----- Message d'origine -----
De=20 : Johanna =
Envoy=E9 : Thursday, = September 25,=20 2003 9:46 AM
Objet : [MT-List] MT on-line=20 services

My name is Johanna Monti and I teach = History and=20 Theory of Translation at the University L'Orientale of Neaples. =
I'm looking for the following=20 infomation:
how many users use MT on-line = services=20 ?
how many words/lines, etc. are = translated per=20 day/year by MT on-line services?
what type of texts are translated and = in which=20 percentage?
what is the purpose of the = translation (internal=20 use, for information dissemination, etc.)?
 
Does anybody know if there are any=20 articles/figures available about this issue?
 
Thanks for your help.
 
Johanna=20 Monti
------=_NextPart_000_0362_01C385DD.AF9BDD40-- From jeff.allen@free.fr Sun Sep 28 21:14:01 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Sun, 28 Sep 2003 22:14:01 +0200 Subject: [MT-List] Pricing Post-editing of MT References: <22772E6015DA354B84C6F8DB8BF60504093344C9@dewdfx13.wdf.sap.corp> Message-ID: <05bc01c38606$d6223e40$449a933e@home> > >>> wrote:>>> > I'm looking for information about the prices of post-editing of raw MT output. > Whereas human translation is mainly based on the unit "word" as a cost base, > in the case of the post-editor, which is the situation? how much is s/he > paid, more or less? Lorena, sorry for my late reply. see: Postediting at ABLE International (cf. http://linguistlist.org/issues/10/10-1258.html) Since 1996, I have noticed a trend among translation agencies and free-lance translators moving toward per hour fees for both TM fuzzy-matching editing and MT postediting tasks. > Henrik Gade [mailto:henrik.wilhelm.gade@medtronic.com] wrote: > We regard post-editing as a qualified human translation work with a little MT help to speed up the process, We are producing literally hundreds of manuals a year, so speed is essential. We don't use fuzzy matches - Systranet produces really good raw texts, if we fine tune the English original. Henrik, If you are pre-editing the text (ie controlling the input text), and the texts have any significant amount of repetition, then your re-authoring / translation cycle could be optimized. Why not apply memory-based strategies to the input so that you can better leverage the texts? See: Adapting the concept of "Translation Memory" to "Authoring memory" for a Controlled Language writing environment. (Available on-line at: http://www.transref.org/u-articles/allen2.asp) > >>> "Grasmick, Daniel" wrote>>> > how do you reward the translators doing post-editing of TM fuzzy matches? I would say to give them translator-friendly software that corresponds to their needs. For example, providing the output in SGML or XML tagged texts with font differentiated tags that separate the TM fuzzy match output from MT output and make it easier to identify them. Also, translation-specific features like a word mover (see description of such a feature in: http://www.multilingual.com/allen50.htm ) Placing MT output into MS Word files and using Word to postedit the documents is inefficient use of a translator's time unless Word has been customized to offer the same kinds of editing features that are found in the desktop interface versions of translation software. Regards, Jeff From nadamides@aslib.com" Dear Colleague, Translating and the Computer 25 - Conference and Exhibition Thursday 20 - Friday 21 November 2003 at the Commonwealth Institute, London The Early Bird Discount deadline for this event expires tomorrow (1st October) - book in as soon as possible to ensure your place! Members of EAMT and IAMT are offered the member's rate for the conference. A full programme and other details can be found at: www.aslib.com/conferences/#TC25 Contributors include: - Sylvia Ball, European Parliament - Pernilla Danielsson, University of Birmingham, UK - Yves Champollion, France - Dr Imelda P. de Castro, AB Translation Services, De La Salle University, Manila - Daniel Gervais, MultiCorpora R&D Inc., Canada - Luc Huygh, Euroscript, Luxembourg - JosuKa Diaz Labrador, Universidad de Deusto, Spain - Dr Minako O'Hagan, Dublin City University, Ireland - Professor Reinhard Schaler, University of Limerick, Ireland - Adriane Rinsche, Language Technology Centre, Ltd., UK - Anja Rutten, Conference Interpreter, Germany - Ross Smith, Translation Service, PricewaterhouseCoopers, Spain - Jaap van der Meer, Cross Language, Belgium - Andrzej Zydron, UK The conference is supported by BCS, EAMT, IAMT, ITI and IoL. If you have any questions or would like more information, please do not hesitate to contact me. Nicole Adamides, ASLIB Training Aslib/IMI, Temple Chambers, 3-7 Temple Avenue, London, EC4Y 0HP Tel: +44 (0)20 7583 8900 Fax: +44 (0)20 7583 8401 www.aslib.com From jeff.allen@free.fr Mon Oct 13 13:38:25 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Mon, 13 Oct 2003 14:38:25 +0200 Subject: [MT-List] EAMT/CLAW2003 presentations Message-ID: <1066048705.3f8a9cc1dda59@imp3-l.free.fr> Dear all, Follow-up info on the EAMT/CLAW2003 Controlled Language Translation Conference (May 2003) can be accessed online: Summary of all past EAMT and CLAW conferences http://www.multilingual.com/allen54.htm Follow-up event article on EAMT/CLAW2003 http://www.dcu.ie/news/archive/may03/s0305g.shtml EAMT/CLAW2003 Conference program http://www.eamt.org/eamt-claw03/programme.html EAMT/CLAW2003 Conference presentations http://www.ctts.dcu.ie/presentations.html regards, Jeff Allen jeff.allen@free.fr From anneleenpareyn@hotmail.com Mon Oct 13 14:45:43 2003 From: anneleenpareyn@hotmail.com (Anneleen Pareyn) Date: Mon, 13 Oct 2003 15:45:43 +0200 Subject: [MT-List] Controlled Language Message-ID: Dear all, this year I have to write my dissertation. The purpose is to compare a translation by a human translator (that'll be me, as I intend to graduate as a translator) and two machine translation systems. I also have to write some chapters about MT and everything that has to do with it. One of the subjects will be Controlled Language. My question now is: Is Controlled Language really necessary to have a "perfect" MT output? Are there any books or special articles about this subject? Thanks, Anneleen _________________________________________________________________ Chatten met je online vrienden via MSN Messenger. http://messenger.msn.be From andrei.popescu-belis@issco.unige.ch Mon Oct 13 15:42:30 2003 From: andrei.popescu-belis@issco.unige.ch (Andrei Popescu-Belis) Date: Mon, 13 Oct 2003 16:42:30 +0200 Subject: [MT-List] Framework for MT Evaluation (FEMTI) Message-ID: <3F8AB9D6.1000807@issco.unige.ch> Dear all, We are pleased to announce that FEMTI, the Framework for Machine Translation Evaluation in ISLE, is available for consultation at the following mirrored URLs: http://www.isi.edu/natural-language/mteval/ http://www.issco.unige.ch/projects/isle/femti/ FEMTI is a structured repertoire of methods used to evaluate MT systems. The first part of FEMTI enables evaluators to specify a context of use for the MT system to be evaluated, while the second part defines relevant quality characteristics and metrics. FEMTI is a result of the ISLE Project, funded by the US National Science Foundation and the Swiss and Danish Governments. A series of workshops provided valuable input to the framework (see http://www.issco.unige.ch/projects/isle/ewg.html ). The framework is discussed in the following article: Eduard Hovy, Margaret King and Andrei Popescu-Belis (2002) - "Principles of Context-Based Machine Translation Evaluation", Machine Translation, 17:1, p.43-75, 2002. Users are welcome to browse the framework at the above URLs and to use the feedback buttons to provide comments. The FEMTI coordinators: Eduard Hovy Margaret King Andrei Popescu-Belis ----------------------- From jeff.allen@free.fr Mon Oct 13 17:52:03 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Mon, 13 Oct 2003 18:52:03 +0200 Subject: [MT-List] Controlled Language In-Reply-To: References: Message-ID: <1066063923.3f8ad8334861f@imp3-l.free.fr> Aneeleen, You can start by looking at all the MT-List postings below which will provide all the additional links to information on the topic: http://www.mail-archive.com/mt-list@eamt.org/msg00037.html http://www.mail-archive.com/mt-list@eamt.org/msg00120.html http://www.mail-archive.com/mt-list@eamt.org/msg00399.html http://www.mail-archive.com/mt-list@eamt.org/msg00453.html http://www.mail-archive.com/mt-list@eamt.org/msg00487.html http://www.mail-archive.com/mt-list@eamt.org/msg00578.html http://www.mail-archive.com/mt-list@eamt.org/msg00579.html http://www.mail-archive.com/mt-list@eamt.org/msg00581.html http://www.mail-archive.com/mt-list@eamt.org/msg00549.html http://www.mail-archive.com/mt-list@eamt.org/msg00550.html http://www.mail-archive.com/mt-list@eamt.org/msg00599.html Also see: CLAW96 http://www.ccl.kuleuven.ac.be/CLAW/programme.html CLAW98 http://www.lti.cs.cmu.edu/CLAW98/ CLAW2000 http://www.up.univ-mrs.fr/~veronis/claw2000/ EAMT/CLAW2003 http://www.eamt.org/eamt-claw03/index.html Conference presentations http://www.ctts.dcu.ie/presentations.html CLAW and EAMT conferences http://www.multilingual.com/allen54.htm Archives of past EAMT workshops http://www.eamt.org/archive.html Regards, Jeff Allen Quoting Anneleen Pareyn : > Dear all, > > this year I have to write my dissertation. The purpose is to compare a > translation by a human translator (that'll be me, as I intend to graduate as > > a translator) and two machine translation systems. > I also have to write some chapters about MT and everything that has to do > with it. One of the subjects will be Controlled Language. > > My question now is: Is Controlled Language really necessary to have a > "perfect" MT output? > Are there any books or special articles about this subject? > > Thanks, > Anneleen > -- From away@computing.dcu.ie Tue Oct 14 16:29:55 2003 From: away@computing.dcu.ie (Andy Way) Date: Tue, 14 Oct 2003 16:29:55 +0100 Subject: [MT-List] EAMT-CLAW 2003 Proceedings available Message-ID: <3F8C1673.6000009@computing.dcu.ie> Hi all, I thought I'd sent this a week or so ago, but didn't see it in the digest, hence the reposting. Anyway, just to say that copies of the proceedings of EAMT-CLAW 03 can be ordered on-line at: http://www.eamt.org/eamt-claw03/proceedingsorderform.html You get both the printed volume and CD-ROM for your ¤20. Andy. PS, Please circulate to interested parties at AMTA and AAMT. Thanks! From rasmusse@research.rutgers.edu Mon Oct 20 17:30:18 2003 From: rasmusse@research.rutgers.edu (Priscilla Rasmussen) Date: Mon, 20 Oct 2003 12:30:18 EDT Subject: [MT-List] MT Summit IX Conference Presentations/Papers Now Available Online Message-ID: MT SUMMIT IX Conference Proceedings Are Now Available Online! The conference proceedings of the recently held MT Summit IX in New Orleans are now available online. These pages contain links to the files of nearly all the papers and presentations given in New Orleans and contain: -Preface by the conference chair (EduardHovy). -Preface by the program chair (Elliott Macklovitch). -Invited Speakers presentations or file notes. -Panel on "Have we found the Holy Grail?" and questions asked of the panelists. -Presentations -Research Papers and User Studies To browse through the activities and publications, pease go to http://www.amtaweb.org/summit/MT Summit/papers.html From Christian.Boitet@imag.fr Mon Nov 3 21:23:52 2003 From: Christian.Boitet@imag.fr (Christian Boitet) Date: Mon, 3 Nov 2003 22:23:52 +0100 Subject: [MT-List] Controlled Language is NOT necessary for FAHQMT In-Reply-To: References: Message-ID: Dear Anneleen, 3/11/03 At 15:45 +0200 13/10/03, Anneleen Pareyn wrote: >Dear all, > >this year I have to write my dissertation. The purpose is to compare >a translation by a human translator (that'll be me, as I intend to >graduate as a translator) and two machine translation systems. >I also have to write some chapters about MT and everything that has >to do with it. One of the subjects will be Controlled Language. > >My question now is: Is Controlled Language really necessary to have >a "perfect" MT output? No! See below. And=8A what is a "perfect" (human or machine) translation output anyway? Did your translation teacher give you some objective criteria for that? >Are there any books or special articles about this subject? > >Thanks, >Anneleen I take it you mean "100% automatic MT output". Then, A RESTRICTED ENOUGH SUBLANGUAGE, NOT CONTROLLED but resulting from some (ever slightly shifting) conventions and more or less formal rules of "well writing" IS ENOUGH TO GET NEAR PERFECT MT OUTPUT --- with a lot of ingenuity and elbow grease, that is. Example: the sublanguage of the weather bulletins handled by METEO is NOT a controlled language. But METEO was last reported to be "97% correct" (in 1985, the figure was "only" 85%), or rather "3% less than perfect" by the measure of how many text editor operations (for 100 words) are performed in average by the human posteditors to get what they judge to be "professional" (perfect?) output. That corresponds to 1 mn for a typical bulletin of 100-150 words or so. By comparison, the same posteditors (I mean, posteditors with the same level of excellence) had to spend about 7-10 mn before the first version of METEO, TAUM-METEO, was introduced (in 1976?). That is because translators producing the draft translations were junior translators assigned to this "purgatory" just a few months, the time to master this difficult kind of translation=8A and to become fed up with it. (By the time it was translated and sent back to wherever it came from, a bulletin had only 2 hours to live before becoming obsolete. Not very motivating!) By that measure, METEO is about 7-10 times better than an average, learning, sweating and swearing junior translator just put on the job. I think such task-oriented metrics are the best one for judging QUALITY-oriented MT, that is, "MT for revisors" as opposed to "MT for watchers". Maybe John Chandioux could correct and complete what I sketched above, as well as give you pointers to some papers on the topic. Best, Ch.Boitet From jeff.allen@free.fr Tue Nov 4 12:50:41 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Tue, 4 Nov 2003 13:50:41 +0100 Subject: [MT-List] re: Controlled Language is NOT necessary for FAHQMT Message-ID: <1067950241.3fa7a0a1393a7@imp3-l.free.fr> Online references that address many of the points in Christian Boitet's posting the MT-List yeserday (pasted in further below) are available at: http://www.mail-archive.com/mt-list%40eamt.org/msg00599.html http://www.benjamins.com/cgi-bin/t_bookview.cgi?bookid=BTL_35 read chapter 16 http://www.ctts.dcu.ie/presentations.html http://www.multilingual.com/allen58.htm http://www.multilingual.com/allen50.htm http://www.multilingual.com/allen46.htm http://www.transref.org/u-articles/allen2.asp http://www.tc-forum.org/topiccl/cl15diff.htm http://www.controlled-language.org Those provide many additional sets of references. Many other references also cited in previous MT-List postings: http://www.mail-archive.com/mt-list%40eamt.org/msg00578.html http://www.mail-archive.com/mt-list%40eamt.org/msg00579.html http://www.mail-archive.com/mt-list%40eamt.org/msg00549.html http://www.mail-archive.com/mt-list%40eamt.org/msg00550.html http://www.mail-archive.com/mt-list%40eamt.org/msg00552.html Regards, Jeff Allen ---------------- De : "Christian Boitet" Ŕ : "Anneleen Pareyn" ; Cc : "J.Chandioux & A.Grimaila" <104213.451@compuserve.com> Envoyé : Monday, November 03, 2003 10:23 PM Objet : [MT-List] Controlled Language is NOT necessary for FAHQMT Dear Anneleen, 3/11/03 At 15:45 +0200 13/10/03, Anneleen Pareyn wrote: >Dear all, > >this year I have to write my dissertation. The purpose is to compare >a translation by a human translator (that'll be me, as I intend to >graduate as a translator) and two machine translation systems. >I also have to write some chapters about MT and everything that has >to do with it. One of the subjects will be Controlled Language. > >My question now is: Is Controlled Language really necessary to have >a "perfect" MT output? No! See below. AndS what is a "perfect" (human or machine) translation output anyway? Did your translation teacher give you some objective criteria for that? >Are there any books or special articles about this subject? > >Thanks, >Anneleen I take it you mean "100% automatic MT output". Then, A RESTRICTED ENOUGH SUBLANGUAGE, NOT CONTROLLED but resulting from some (ever slightly shifting) conventions and more or less formal rules of "well writing" IS ENOUGH TO GET NEAR PERFECT MT OUTPUT --- with a lot of ingenuity and elbow grease, that is. Example: the sublanguage of the weather bulletins handled by METEO is NOT a controlled language. But METEO was last reported to be "97% correct" (in 1985, the figure was "only" 85%), or rather "3% less than perfect" by the measure of how many text editor operations (for 100 words) are performed in average by the human posteditors to get what they judge to be "professional" (perfect?) output. That corresponds to 1 mn for a typical bulletin of 100-150 words or so. By comparison, the same posteditors (I mean, posteditors with the same level of excellence) had to spend about 7-10 mn before the first version of METEO, TAUM-METEO, was introduced (in 1976?). That is because translators producing the draft translations were junior translators assigned to this "purgatory" just a few months, the time to master this difficult kind of translationS and to become fed up with it. (By the time it was translated and sent back to wherever it came from, a bulletin had only 2 hours to live before becoming obsolete. Not very motivating!) By that measure, METEO is about 7-10 times better than an average, learning, sweating and swearing junior translator just put on the job. I think such task-oriented metrics are the best one for judging QUALITY-oriented MT, that is, "MT for revisors" as opposed to "MT for watchers". Maybe John Chandioux could correct and complete what I sketched above, as well as give you pointers to some papers on the topic. Best, Ch.Boitet From hannouna@uruklink.net Thu Nov 6 09:07:28 2003 From: hannouna@uruklink.net (hannouna) Date: Thu, 6 Nov 2003 12:07:28 +0300 Subject: [MT-List] Controlled language Message-ID: <00d801c3a445$67985100$090a010a@yasmin> This is a multi-part message in MIME format. ------=_NextPart_000_00D4_01C3A45E.8C1EDDA0 Content-Type: multipart/alternative; boundary="----=_NextPart_001_00D5_01C3A45E.8C1EDDA0" ------=_NextPart_001_00D5_01C3A45E.8C1EDDA0 Content-Type: text/plain; charset="windows-1256" Content-Transfer-Encoding: quoted-printable BlankDear Anneleen, Hello. If you want to compare your translation to MT output produced by certain systems , it means that you have to deal with the Evaluation of = MT which is a special field of knowledge and research by its own and there = is a huge literature on this subject . I suggest you can start with the ISLE Framework , (better called 'FEMTI'). The result (taxonomy3) is now available at: > http://www.issco.unige.ch/projects/isle/taxonomy3/ > (better for you to download the printable version at:http://www.issco.unige.ch/projects/isle/taxonomy3/printableclassificat= ion .ht Where you can find and select the appropriate quality criteria for your evaluation. As for a controlled language , it is referred to as a special simplified version of a language which is adopted ( typically by a company or a documentation section of a company ) as a partial solution to a = perceived communication problem . Both the vocabulary and the syntactic structure = may be restricted . You can find more discussion on CL in any recent reference on MT or = through online search ( e.g. Google ) . I think it is useful for a better MT = output. If you need any further information , please contact me. Best wishes. Yasmin Hannouna. ----- Original Message ----- From: "Christian Boitet" To: "Anneleen Pareyn" ; Cc: "J.Chandioux & A.Grimaila" <104213.451@compuserve.com> Sent: Tuesday, November 04, 2003 12:23 AM Subject: [MT-List] Controlled Language is NOT necessary for FAHQMT Dear Anneleen, 3/11/03 At 15:45 +0200 13/10/03, Anneleen Pareyn wrote: >Dear all, > >this year I have to write my dissertation. The purpose is to compare >a translation by a human translator (that'll be me, as I intend to >graduate as a translator) and two machine translation systems. >I also have to write some chapters about MT and everything that has >to do with it. One of the subjects will be Controlled Language. > >My question now is: Is Controlled Language really necessary to have >a "perfect" MT output? No! See below. AndS what is a "perfect" (human or machine) translation output anyway? Did your translation teacher give you some objective criteria for that? >Are there any books or special articles about this subject? > >Thanks, >Anneleen I take it you mean "100% automatic MT output". Then, A RESTRICTED ENOUGH SUBLANGUAGE, NOT CONTROLLED but resulting from some (ever slightly shifting) conventions and more or less formal rules of "well writing" IS ENOUGH TO GET NEAR PERFECT MT OUTPUT --- with a lot of ingenuity and elbow grease, that is. Example: the sublanguage of the weather bulletins handled by METEO is NOT a controlled language. But METEO was last reported to be "97% correct" (in 1985, the figure was "only" 85%), or rather "3% less than perfect" by the measure of how many text editor operations (for 100 words) are performed in average by the human posteditors to get what they judge to be "professional" (perfect?) output. That corresponds to 1 mn for a typical bulletin of 100-150 words or so. By comparison, the same posteditors (I mean, posteditors with the same level of excellence) had to spend about 7-10 mn before the first version of METEO, TAUM-METEO, was introduced (in 1976?). That is because translators producing the draft translations were junior translators assigned to this "purgatory" just a few months, the time to master this difficult kind of translationS and to become fed up with it. (By the time it was translated and sent back to wherever it came from, a bulletin had only 2 hours to live before becoming obsolete. Not very motivating!) By that measure, METEO is about 7-10 times better than an average, learning, sweating and swearing junior translator just put on the job. I think such task-oriented metrics are the best one for judging QUALITY-oriented MT, that is, "MT for revisors" as opposed to "MT for watchers". Maybe John Chandioux could correct and complete what I sketched above, as well as give you pointers to some papers on the topic. Best, Ch.Boitet -- For MT-List info, see http://www.eamt.org/mt-list.html ------=_NextPart_001_00D5_01C3A45E.8C1EDDA0 Content-Type: text/html; charset="windows-1256" Content-Transfer-Encoding: quoted-printable Blank
Dear Anneleen,

Hello. If you want to compare your = translation to MT=20 output produced by
certain systems , it means that you have to deal = with the=20 Evaluation of MT
which is a special field of knowledge and research = by its=20 own and there is a
huge literature on this subject . I suggest you = can start=20 with the ISLE
Framework ,  (better called 'FEMTI'). The result=20 (taxonomy3) is now
available at:
> http://www.is= sco.unige.ch/projects/isle/taxonomy3/
>=20 (better for you to download the printable=20 version
at:http://www.issco.unige.ch/projects/isle/taxonomy3/printable= classification
.ht

Where=20 you can find and select the appropriate quality criteria for=20 your
evaluation.

As for a controlled language , it is referred = to as a=20 special simplified
version of a language which is adopted ( typically = by a=20 company or a
documentation section of a company ) as a partial = solution to a=20 perceived
communication problem . Both the vocabulary and the = syntactic=20 structure may
be restricted .

You can find more discussion on = CL in=20 any recent reference on MT or through
online search ( e.g. Google ) . = I think=20 it is useful for a better MT output.

If you need any further = information=20 , please contact me.

Best wishes.

Yasmin=20 Hannouna.


----- Original Message -----
From: "Christian = Boitet"=20 <Christian.Boitet@imag.fr>=
To:=20 "Anneleen Pareyn" <anneleenpareyn@hotmail.com= >;=20 <mt-list@eamt.org>
Cc:=20 "J.Chandioux & A.Grimaila" <104213.451@compuserve.com&g= t;
Sent:=20 Tuesday, November 04, 2003 12:23 AM
Subject: [MT-List] Controlled = Language is=20 NOT necessary for FAHQMT


Dear Anneleen, 3/11/03

At = 15:45 +0200=20 13/10/03, Anneleen Pareyn wrote:
>Dear all,
>
>this = year I=20 have to write my dissertation. The purpose is to compare
>a = translation by=20 a human translator (that'll be me, as I intend to
>graduate as a=20 translator) and two machine translation systems.
>I also have to = write=20 some chapters about MT and everything that has
>to do with it. One = of the=20 subjects will be Controlled Language.
>
>My question now is: = Is=20 Controlled Language really necessary to have
>a "perfect" MT=20 output?

No! See below.

AndS what is a "perfect" (human or = machine)=20 translation output
anyway? Did your translation teacher give you some = objective criteria
for that?

>Are there any books or = special=20 articles about this = subject?
>
>Thanks,
>Anneleen

I=20 take it you mean "100% automatic MT output".

Then, A RESTRICTED = ENOUGH=20 SUBLANGUAGE, NOT CONTROLLED but resulting
from some (ever slightly = shifting)=20 conventions and more or less
formal rules of "well writing" IS ENOUGH = TO GET=20 NEAR PERFECT MT
OUTPUT --- with a lot of ingenuity and elbow grease, = that=20 is.

Example: the sublanguage of the weather bulletins handled by = METEO=20 is
NOT a controlled language. But METEO was last reported to be=20 "97%
correct" (in 1985, the figure was "only" 85%), or rather "3%=20 less
than perfect" by the measure of how many text editor operations=20 (for
100 words) are performed in average by the human posteditors to=20 get
what they judge to be "professional" (perfect?) output.=20 That
corresponds to 1 mn for a typical bulletin of 100-150 words or=20 so.

By comparison, the same posteditors (I mean, posteditors with = the
same level of excellence) had to spend about 7-10 mn before the=20 first
version of METEO, TAUM-METEO, was introduced (in 1976?). That=20 is
because translators producing the draft translations were=20 junior
translators assigned to this "purgatory" just a few months, = the=20 time
to master this difficult kind of translationS and to become fed=20 up
with it. (By the time it was translated and sent back to wherever=20 it
came from, a bulletin had only 2 hours to live before=20 becoming
obsolete. Not very motivating!)

By that measure, = METEO is=20 about 7-10 times better than an average,
learning, sweating and = swearing=20 junior translator just put on the job.

I think such task-oriented = metrics=20 are the best one for judging
QUALITY-oriented MT, that is, "MT for = revisors"=20 as opposed to "MT for
watchers".

Maybe John Chandioux could = correct=20 and complete what I sketched
above, as well as give you pointers to = some=20 papers on the = topic.

Best,

Ch.Boitet



--
 =20 For MT-List info, see http://www.eamt.org/mt-list.htm= l

 

------=_NextPart_001_00D5_01C3A45E.8C1EDDA0-- ------=_NextPart_000_00D4_01C3A45E.8C1EDDA0 Content-Type: image/gif; name="Blank Bkgrd.gif" Content-Transfer-Encoding: base64 Content-ID: <00d301c3a445$66ce9860$090a010a@yasmin> R0lGODlhLQAtAID/AP////f39ywAAAAALQAtAEACcAxup8vtvxKQsFon6d02898pGkgiYoCm6sq2 7iqWcmzOsmeXeA7uPJd5CYdD2g9oPF58ygqz+XhCG9JpJGmlYrPXGlfr/Yo/VW45e7amp2tou/lW xo/zX513z+Vt+1n/tiX2pxP4NUhy2FM4xtjIUQAAOw== ------=_NextPart_000_00D4_01C3A45E.8C1EDDA0-- From malek.boualem@rd.francetelecom.com Thu Nov 6 09:29:02 2003 From: malek.boualem@rd.francetelecom.com (BOUALEM Malek FTRD/DMI/LAN) Date: Thu, 6 Nov 2003 10:29:02 +0100 Subject: [MT-List] CFP - JEP-TALN-2004 session on Arabic Language Processing Message-ID: <749A556ABA92204FA7E566D8CA6A169B5CC708@ftrdmel1.rd.francetelecom.fr> ******************************************************** J E P 2 0 0 4 - T A L N 2 0 0 4 - Special Session - --- ARABIC LANGUAGE PROCESSING =20 Call for Papers --- Palais des Congr=E8s Fez (Morocco) 19-22 April 2004 http://www.lpl.univ-aix.fr/jep-taln04/ http://www.fsdmfes.ac.ma/jep-taln04/ ******************************************************** Due to its morphological, syntactic, phonetic and phonologic=20 properties, the Arabic language is considered to be one of the most difficult languages for written and spoken language processing. Research on written Arabic language processing started in the 1970s, even before the problems of Arabic text editing were completely solved. The first studies focused primarily on lexicons and morphology. In the past ten years, the internationalisation of the WWW and the proliferation of communication tools in Arabic have led to the need for a large number of Arabic NLP applications. As a result, research activity has extended to address more general areas of Arabic language processing, including syntactic analysis, machine translation, document indexing, information retrieval, etc. Research on Arabic speech processing has made significant progress due to more improved signal processing technologies, and to recent advances in the knowledge of the prosodic and the segmental characteristics of Arabic and the acoustic modelling of Arab schemes. These results should make it possible to further progress in more innovative areas, such as Arabic speech recognition and synthesis, speech translation and automatic identification of a speaker and his/her geographic origin discrimination, etc. The aim of the joint session is to gather and reinforce collaboration between researchers from both the written and spoken Arabic language processing communities. It will also offer the opportunity to discuss recent advances on both the scientific and application sides of the problem, in monolingual and multilingual contexts. TOPICS This special session on written and spoken Arabic processing includes (but is not limited to) the following topics : - Speech recognition and comprehension, - Text to speech synthesis, - Automatic prosody generation, - Automatic speaker and language identification, - Geographic origin discrimination of Arabic speakers, - Arabic corpora & resources, - Speech acquisition for ASR and TTS systems, - Morphology, - Syntax, - Semantics, - Text parsing and generation, - Discourse analysis, - Text summarization, - Dialogue, - Machine translation. IMPORTANT DATES Submission deadline : 15 January 2004=20 Notification to authors : 20 February 2004=20 Camera-ready : 8 March 2004=20 Conference : 19-22 April 2004 **** only for the PC members ***** According to the detailed calendar below : - Submission deadline: 15 January 2004 - Submissions sent to the reviewers : 22 January 2004 - Reviews ready and sent back : 13 February 2004 - Notification to authors: 20 February 2004=20 - Corrections sent back : 27 February 2004 - Corrections validation : 05 March 2004 - Camera-ready: 8 March 2004 - Conference: 19-22 April 2004 **** only for the PC members ***** =20 SELECTION, LANGUAGES AND SUBMISSION PROCEDURE Please check the available information on the conference web site : http://www.lpl.univ-aix.fr/jep-taln04/ SUBMISSION ADDRESS Electronic submissions, with the message object "JEP-TALN-2004-Arabic", have to be sent to the following email address : < jep-taln04-arabic@fsdmfes.ac.ma > In case electronic submissions are not possible, printed versions might be accepted. In this case, three hard-copies of the paper together with a floppy disk, have to be sent to : Malek Boualem France Telecom R&D - DMI/GRI 2, avenue Pierre Marzin 22307 Lannion - France or=20 Noureddine Chenfour D=E9partement de Math. et Informatique Facult=E9 des Sciences Dhar El Mahraz, F=E8s=20 BP : 1796 Atlas, F=E8s - Maroc ******************************************************** ------------------------------------------------------ Malek Boualem France Telecom R&D - DMI/GRI 2, avenue Pierre Marzin - 22307 Lannion - France Tel: (33)(0)2.96.05.29.83 Fax: (33)(0)2.96.05.32.86 Email: malek.boualem@rd.francetelecom.com ------------------------------------------------------ From jeff.allen@free.fr Thu Nov 6 12:42:08 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Thu, 6 Nov 2003 13:42:08 +0100 Subject: [MT-List] parallel texts online Message-ID: <1068122528.3faa41a03e8d8@imp2-l.free.fr> Someone made a post to the list a few months ago about where to find parallel texts. The MT archives site does not have original post nor the summary post, so I don't remember who to send it to directly. See the Euromap article archives for several articles available in bilingual format. http://www.hltcentral.org/page-960.shtml regards, Jeff From cb@lim.nl Sat Nov 8 11:37:14 2003 From: cb@lim.nl (Colin Brace) Date: Sat, 08 Nov 2003 12:37:14 +0100 Subject: [MT-List] parallel texts online In-Reply-To: <1068122528.3faa41a03e8d8@imp2-l.free.fr> Message-ID: On 11/06/03 at 01:42 PM, Jeff Allen said: > Someone made a post to the list a few months ago about where to find > parallel texts. The MT archives site does not have original post nor > the summary post, so I don't remember who to send it to directly. The message was from D Elliott on 21 May 2003, subject: Parallel texts for machine translation evaluation Indeed, for some inexplicable reason, the list archive at mail-archive.com has a lacuna of several weeks in May, but I found the message in the pairlist.net archive: http://www.pairlist.net/pipermail/mt-list/2003/000286.html replies can be viewed here: http://www.pairlist.net/pipermail/mt-list/2003/thread.html#286 Yes, for a variety of complicated and not terribly interesting reasons, there are two archives: http://www.pairlist.net/pipermail/mt-list/ http://www.mail-archive.com/mt-list%40eamt.org/ Sometime ago I updated the list info page to indicate this but forgot to add the information to http://www.eamt.org/mt-list.html -- Colin Brace | cb@lim.nl EAMT Webmaster From dave@lsbu.ac.uk Tue Nov 11 10:19:10 2003 From: dave@lsbu.ac.uk (DaveInman) Date: Tue, 11 Nov 2003 10:19:10 +0000 Subject: [MT-List] Adjective noun frequencies Message-ID: <5.1.0.14.0.20031111100852.00b8d590@mail.onetel.net.uk> I need to find adjective-noun frequencies for English. Any leads? Dave Inman, London South Bank University. From majithia@streamsage.com Thu Nov 13 21:06:01 2003 From: majithia@streamsage.com (Hemali Majithia) Date: Thu, 13 Nov 2003 16:06:01 -0500 (EST) Subject: [MT-List] core dump in GIZA++ In-Reply-To: References: Message-ID: Hi all, I am running GIZA++ with a 10MB of sample corpus (from the European Parliament). It has around 30K sentence pairs. The code core dumps (segmentation fault) when it's starts training for the HMM model. From what I understand, this happens in the em_loop. I don't face this problem when I try to train it for a smaller corpus size (1MB). I was hoping to know if anybody has run into a similar problem and if so, is there something that I should be doing. I run it on a linux machine with 1GB of memory (so it's sufficiently large). Thanks, Hemali From debe@comp.leeds.ac.uk Thu Nov 27 16:17:17 2003 From: debe@comp.leeds.ac.uk (D Elliott) Date: Thu, 27 Nov 2003 16:17:17 +0000 (GMT) Subject: [MT-List] AMTA 1994 proceedings? Message-ID: I am desperately trying to get hold of the following papers from the first AMTA Conference 1994. Unfortunately these don't appear to be available online (unlike papers from subsequent conferences). If anyone can help I'd be very grateful. Flanagan, M. Error Classification for MT Evaluation. In Technology Partnerships for Crossing the Language Barrier: Proceedings of the First Conference of the Association for Machine Translation in the Americas, (Columbia, Md. 1994). White, J.S., O'Connell, T., and O'Mara, F.E. The ARPA MT Evaluation Methodologies: Evolution, Lessons and Further Approaches. In Technology Partnerships for Crossing the Language Barrier: Proceedings of the First Conference of the Association for Machine Translation in the Americas. (Columbia, Md. 1994.) Thanks, Debbie Elliott -- *************************************************** Debbie Elliott Computer Vision and Language Research Group, School of Computing, University of Leeds, Leeds LS2 9JT United Kingdom. Tel: 0113 3436818 Email: debe@comp.leeds.ac.uk *************************************************** From WJHutchins@compuserve.com Mon Dec 1 17:55:43 2003 From: WJHutchins@compuserve.com (John Hutchins) Date: Mon, 1 Dec 2003 12:55:43 -0500 Subject: [MT-List] EAMT workshop Malta, April 2004 Message-ID: <200312011256_MC3-1-5D34-29D2@compuserve.com> This is a MIME-encapsulated message --d2c4ff83-8988-464c-9b02-d2eb4ff718d7 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Content-Disposition: inline EAMT Workshop Malta, 26-27 April 2004 We are sending out the first call for papers. Please see the attached file for details. John Hutchins 1 Dec --d2c4ff83-8988-464c-9b02-d2eb4ff718d7 Content-Type: text/plain; charset=ISO-8859-1; name="EAMT Workshop 2004 CFP.txt" Content-Transfer-Encoding: quoted-printable Content-Disposition: attachment; filename="EAMT Workshop 2004 CFP.txt" European Association for Machine Translation Workshop, Malta, 26-27 April 2004 First call for papers The European Association for Machine Translation (EAMT) will be holding i= ts = Ninth Workshop in Malta on 26-27 April 2004. The meeting is one of a seri= es of = annual workshops acting as a forum for the exchange of ideas concerning a= ll = aspects of machine translation within the European and adjacent regions.= (For = information about previous workshops see the EAMT website: www.eamt.org.)= As in = previous EAMT workshops, contributions will be particularly welcomed from= = developers, vendors and users of MT systems and translation aids. = The SPECIAL THEME for this workshop will be machine-translation-related i= ssues = concerning = = * Semitic languages, * Languages of the newly accessioned states of the European Union. Papers on machine translation systems and computer-based translation aids= = relating to these languages will be of particular interest. = However, papers relating to other aspects of machine translation, and oth= er = languages, will also be considered such as: * Translation Memories * Web-based MT services = * Issues for developers and users of MT and translation aids = * Economic aspects of MT * Evaluation of MT systems = SUBMISSION of papers: We invite submissions as either extended abstracts (minimum 2 pages) or a= s full = papers (maximum 10 pages). Submissions should be sent before 25th January= 2004, = by email as attachments in DOC, RTF, or plain text, to the chairman John = Hutchins (WJHutchins@compuserve.com), with the words 'EAMT 2004 Workshop'= in the = subject heading). = Notification of acceptance will be made on or before 15 February, togethe= r with = any comments from reviewers for revision. FINAL VERSIONS of papers Final versions of papers (maximum 12 pages, DOC or RTF files) should be s= ent for = inclusion in the proceedings before 12th March. (Any papers received afte= r that = date will not be included in the printed proceedings). A stylesheet for a= ccepted = submissions will be available in due course on the EAMT website. It is anticipated that papers will be made available in electronic form o= n the = EAMT website after the conference -- for this purpose, authors may also s= end PDF = or PS versions of their papers. PROGRAMME COMMITTEE: John Hutchins (chair, EAMT president) Michael Rosner (University of Malta) Anthony Clarke (CLS Corporate Language Services) Jaro Lajovic (Ljubljana, Slovenia) Mustafa Yaseen (ATS Online) Jan Hajic (Charles University, Prague) Violetta Cavalli-Sforza (San Francisco State University) VENUE, registration: Details about the venue in MALTA, registration fees, hotel accommodation,= etc. = will be circulated from mid January or early February. Full details, incl= uding a = list of all speakers and an online registration form, will be posted on t= he EAMT = website (www.eamt.org) in early March. Local arrangements organiser is Mi= ke = Rosner (mike.rosner@um.edu.mt) from whom further details can be obtained.= IMPORTANT DATES: Deadline for submissions: 25 January 2004 Acceptance notification: 15 February 2004 Final copies due: 12 March 2004 Registration: early March 2004 Conference dates: 26-27 April 2004 --d2c4ff83-8988-464c-9b02-d2eb4ff718d7-- From jeff.allen@free.fr Mon Dec 1 12:36:58 2003 From: jeff.allen@free.fr (Jeff Allen) Date: Mon, 1 Dec 2003 13:36:58 +0100 Subject: [MT-List] MT thesis and article online Message-ID: <048301c3b808$0608e040$7a27e4d5@home> C'est un message de format MIME en plusieurs parties. ------=_NextPart_000_047B_01C3B810.31B8FBA0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear all, Some of the references below are now available with direct links: GUERRA, Lorena. 2003. Human Translation versus Machine Translation and = Full Post-Editing of Raw Machine Translation Output. Master's Thesis. = Dublin City University. Completed in August 2003, and available online = November 2003. http://www.promt.ru/news-e/news.phtml?id=3D293 ALLEN, Jeff. 2003. Review of Systran Premium 4.0. Now online at: http://www.multilingual.com/allen58.htm Best regards, Jeff Allen Multilingual Computing and Technology, Editorial Board http://www.multilingual.com/editorialBoard/ ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Cc : jeff.allen@free.fr=20 Envoy=E9 : Saturday, September 27, 2003 4:18 PM Objet : [MT-List] more MT software review articles Dear all, =20 A few software review articles have appeared lately:=20 GUERRA, Lorena. 2003. Review: Machine Translation with @promt = Professional. Irish Translators' & Interpreters' Association (ITIA) = Bulletin. September 2003. pp. 5-6. WASSMER, Thomas. 2003. Review: Systran 4.0: Personal, Standard, = Premium. In Multilingual Computing and Technology magazine. Number 58, = Vol. 14, Issue 6. September 2003. Pp. 17-19. This should be available online at: http://www.multilingual.com/ ALLEN, Jeff. 2003. Review: Systran Premium 4.0. In Multilingual = Computing and Technology magazine. Number 58, Vol. 14, Issue 6. = September 2003. Pp. 19-22. This should be available online at: http://www.multilingual.com/ Also, a thesis/dissertation: GUERRA, Lorena. 2003. Human Translation versus Machine Translation and = Full Post-Editing of Raw Machine Translation Output. Master's = Dissertation. Dublin City University. September 2003.=20 Regards, Jeff --------------------------------- Jeff ALLEN Paris, France e-mail: jeff.allen@free.fr =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Saturday, April 19, 2003 5:02 PM Objet : [MT-List] another MT software review article ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Wednesday, April 02, 2003 7:47 AM Objet : [MT-List] more articles in MLCT on MT and translation = technologies =20 ----- Message d'origine -----=20 De : Jeff Allen=20 =C0 : mt-list@eamt.org=20 Envoy=E9 : Friday, March 14, 2003 10:40 PM Objet : [MT-List] updated links: articles in MLCT on MT, TM, = postediting ------=_NextPart_000_047B_01C3B810.31B8FBA0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable