为了正常的体验网站,请在浏览器设置里面开启Javascript功能!

【计算语言学概论】8 机器翻译

2017-08-14 27页 ppt 128KB 35阅读

用户头像

is_001154

暂无简介

举报
【计算语言学概论】8 机器翻译MachinetranslationIntroductiontoComputationalLinguistics–25April2017IntroductionTranslationoffulltextsfromthesourcelanguagetothetargetlanguageComputerAidedTranslation(CAT)Hardtask:contentshouldbethesamebutthemorphologicalandsyntacticrulesofthetargetlanguagesho...
【计算语言学概论】8 机器翻译
MachinetranslationIntroductiontoComputationalLinguistics–25April2017IntroductionTranslationoffulltextsfromthesourcelanguagetothetargetlanguageComputerAidedTranslation(CAT)Hardtask:contentshouldbethesamebutthemorphologicalandsyntacticrulesofthetargetlanguageshouldbeobservedQualityNowMTqualityiswaybelowthatofhumantranslationBUT:forsomespecific(delimited)topicsitissufficientrecipesweatherforecasttechnicaldescriptionsThemostdesiredNLPapplicationthroughouttheworld„tohelpthosewithdisabilities”Thetaskoftranslation„Ijustlookupthewords,Itranslatethesentence,andit’sdone”Let1sentencecontain15wordsonaverageLet1wordhave3meaningsonaverage1sentencecanhave315differenttranslations*permutationsduetowordorderchanges*insertionordeletionofwordsTheVauqoispyramid InterlinguaSourcelanguagesemantics TargetlanguagesemanticsSourcelanguagesyntax TargetlanguagesyntaxSourcelanguagewords Targetlanguagewords ApproachestoMTDictionarybasedMTSyntactictransformationbasedMTSemantictransformationbasedMTDictionarybasedMTElectronicdictionaryorbilinguallistsofwordsWordsorphrasesinSLarematchedtowordsorphrasesintheTLVerybasicsolutionProblematicissuesarefrequentProblemsinword-to-wordtranslation:Thewordcannotbefoundinthedictionary(SLwordontheTLtext)Toomanysensesinthedictionary(themostfrequentone(?)isselected)Problemsinwordform-to-wordformtranslation:Thedictionarycontainsonlyonewordform(e.g.SgNomfornouns)Inagglutinativelanguages,wordshavemanyforms(HUN)–theycannotbelistedMorphologicalanalysiswouldbenecessaryfortranslationProblemsinphrase-to-phrasetranslation:SyntacticparsingisneededforbothTLandSLResolutionofsyntacticambiguitiesSyntacticparsersshouldbeintegratedSyntactictransformationbasedMTSyntactictreeistransformedintoasyntactictreeNotonlytheleaves(words)aretranslatedThewholetreeistransformedTransformationofatreeOnenodeandonechildisselectedfromthetreewithsomeprobabilityThesenodesareordered(changetheorder,insertionordeletionofothernodes)3.LexicalunitsontheleavesaretranslatedExampleMarydidnotslapthegreenwitchMarianodabaunabofetadaalabrujaverdeCharacteristicsWordorderproblemsmightbesolvedAppropriatesolutionsforrelatedlanguagesComplextransformationrulesIncorrecttranslationsforstructuresthataredifferentinSlandTL(movement+direction):Labotellaentroalacuervaflotando.Thebottlefloatedintothecave.*Thebottleenteredthecavefloating.SemantictransformationSemanticinformationisusedduringMTLinguisticdifferencesamonglanguagescanbeneutralizedSimpleform:applyingsemanticfeatures(human,animate,male/female…)SLinformationistranslatedintoanintermediatelanguage(Interlingua)FrominterlinguaitistranslatedtoTLInformationistranslatednotonlysyntaxNowadays:English~interlinguaExampleJohngaveMaryabook.Give(john;book;mary)T<nowJeanadonnéunlivreàMarie.JánosadottMarinakegykönyvet.SemanticsbasedtranslationSimplerrulesMoreprecisetranslationBasedontheoreticalsemanticsLessexperienceDifficulttoconstructforsomelanguagepairsExamplebasedMT(EBMT)„whenhumanstranslate,theydonotusetransformationalrules,instead,theyfollowpreviouslyseenpatterns”Translationdatabase(parallelcorpus)TranslationunitsareidentifiedThemostsimilarexampleisselectedfromthedatabaseMethodsExamplebased:whichword/phraseismostsimilartotheoneinthedatabaseDictionarybased:weak,maybegoodforrelatedlanguages(syntactic)transferbased:mostmethodsarebasedonthis,acceptablesolutionsInterlingua:translationtointermediatelanguage–„futurework”StatisticalMTEachwordhasseveralpossibletranslationsSelectthemostprobablesequenceofwordsTranslationmodel:bag-of-wordstranslationLanguagemodel:meaningfulsentencesargmaxP(h|e)=argmaxP(h)*P(e|h)BLEU-scoreMeasuringthequalityoftranslationHumanevaluationisverycostlyForthesentencesofthetestset,therearesomehumantranslationsavailableThecoverageofword1,2,3,4gramsiscomparedtothehumantranslationsExampleAFöldnemkör,hanemellipszisalakúpályánkeringaNapkörül.ACsEegységetkorábbanazellipszisfélnagytengelyénekhosszakéntdefiniálták.TheEarthorbitsaroundtheSunonanelliptic,notcircularpath.TheAUwasdefinedashalfthelengthofthemajoraxisoftheellipse.TheEarthrevolvesaroundtheSunnotonacircularbutonanellipticorbit.TheunitAUwasearlierdefinedasthelengthofhalfofthemajoraxisoftheellipse.TheEarth’sorbitaroundtheSunisnotcircular,butelliptical.PreviouslytheAstronomicalUnit(AU)wasdefinedasthelengthofthesemi-majoraxisofthesaidellipse.TheEarthisnotcircular,butellipticalorbitaroundthesun.TheAUunitbeforethesemi-majoraxisoftheellipsewasdefinedasthelength.MTsystemsGoogleTranslate(http://translate.google.com/)Bing(http://www.bing.com/translator)HU-En:MetaMorpho(www.webforditas.hu)computer-aidedtranslation(CAT)tools:IntelligentdictionariesLexicaldatabasesTranslationmemoriesParalellcorporaFunnyexamplesbányászszív [minersucks]bulvárszíndarab [boulevardcolour][piece]gyertyamártás candlesaucehabképző [foam][derivationalsuffix]hajsütés hairbakinghalálnem deathgenderhalmajonéz [dyingmayonnaise]hóhányás [snow][vomit]hóhullás [snow][corpse]hőképzés [heat][training]hőkiütés [heat][knockout]időjóslat [time][prophecy]+lightverbconstruction fényigeszerkezetFunnyexamplesHerchildrengotothesameschoolasmine.Agyerekeibányakéntjárnakugyanabbaaziskolába.(MetaMorpho)Gyermekeimegyugyanaziskolamintbánya.(InterTran)Agyerekeiugyanabbaaziskolábajárnak,mintazenyémek.(reference)FunnyexamplesTherearenobiscuitsleft!Kekszeknincsenekbalra!(MetaMorpho)Nincskétszersültekbal!(InterTran)Nincstöbbkeksz!(reference)
/
本文档为【【计算语言学概论】8 机器翻译】,请使用软件OFFICE或WPS软件打开。作品中的文字与图均可以修改和编辑, 图片更改请在作品中右键图片并更换,文字修改请直接点击文字进行修改,也可以新增和删除文档中的内容。
[版权声明] 本站所有资料为用户分享产生,若发现您的权利被侵害,请联系客服邮件isharekefu@iask.cn,我们尽快处理。 本作品所展示的图片、画像、字体、音乐的版权可能需版权方额外授权,请谨慎使用。 网站提供的党政主题相关内容(国旗、国徽、党徽..)目的在于配合国家政策宣传,仅限个人学习分享使用,禁止用于任何广告和商用目的。

历史搜索

    清空历史搜索