还剩23页未读,继续阅读
本资源只提供10页预览,全部文档请下载后查看!喜欢就下载吧,查找使用更方便
文本内容:
机器翻译与计算机辅助翻译的对比分析semantics hasbegun todeepen inthis period.The basicwork ofcomputational linguistics,such asmany importantalgorithms,has beenthoroughly studied.Lexical analysis andsyntax analysis algorithms have been successivelyconducted,and softwareresources suchas the constructionof electronicdictionaries have been furtherstrengthened.Thetranslation methodis representedby thetransformation method.It startswith theanalysis-based method and theanalysis-based rulemethod,using abstracttransformation torepresentlayered implementationstrategy.Another featureof machine translation atthistime is that thesyntax isseparated fromthe algorithm.The separationof thegrammar fromthealgorithm is to separatethe language analysis fromthe programming.Programmerspropose methodsfor describingrules,and linguistsuse thismethod todescribe therules ofthe language.
2.
1.4New developmentand prosperity1989~presentDuring this period,machine translation has become a hottopic in the study of naturallanguage processing in the world.One of the reasonsis that the demandfor translation isincreasing with the Internetand internationalization,and thetrend ofcommercialization oftranslationsoftware is also extremelyobvious.The translation method inthis periodisgenerally referredto as the thirdgeneration translationmethod,which is mainly based onexamples andstatistics-based methods.The characteristicis to pay attention to theconstructionof large-scale corpusand beginto deal with large-scale realtexts.Anotherimportant featureof thisphase isthat the research workbegins tosolve a more complicatedanddifficult problemthan texttranslation-speech translation.At the same time,since themachine translation systemon Internethas ahuge potentialmarket,the onlinetranslationmachine systemhas alsoentered anew breakthroughstage in the practicalfield.In addition,in thenew phase,machine translation work hasundergone greatchangesin researchand practice.If weput machine translation into the context of natural languageprocessing,I thinkthere arethree mainchanges inthisperiod:First,the combinationoflanguage theoryand computertechnology is more closelyand indepth.Since Chomskypublished“Syntax Structurein1956,linguistics and computer sciencehave combinedtopromote computational linguistics.This is an emergingedge disciplinethat usescomputerto researchand dealwith naturallanguage.Since1980,a large number ofnew grammartheorieshave beenborn,and someof themhavebeenused bylinguists astools fordescribinglanguage phenomena.These grammarsare moreexpressive thanthe simplephrasestructure grammar,which makesit convenientfor linguistsand computerworkers tocommunicate.It alsomakes linguists*latest researchresults morequickly available tocomputer workers.Second,the revivalof statisticalmethods andcorpus methods.Although statisticalmethods do not solveall problems,corpora andstatistical methodshavelargely changedthe faceof machine translation and naturallanguage processing.In thepast,most of the research work wasqualitative research,and quantitativeresearch hasnowbecome mainstream;theresearchobjects have also turnedto large-scale realcorpus.The increasinguse of statistical andregulatory methodsfor machine translation andnaturallanguageprocessinghas also becomeaclear trend.Third,large-scale linguisticdataand testingplatforms havebegun to be applied.Various publiclarge-scale languagematerials,including electronicdictionaries,corpora,and knowledgebases,providefavorable supportfor thestudy of machine translation andnaturallanguageprocessing.Thetraditional machine translation systemmust reachthe initialpractical levelthrough severalyearsor evenmore thanten years of development.It isnow entirelypossible todevelop apreliminaryand practicalmachine translation system in one year.At the same time,theworld hasbegun to adopt a unified testingplatform tobring variousnatural languageprocessingsystems ofdifferent algorithmsdeveloped bydifferent researchinstitutes to aunified environmentfor testing,and toexchange anddiscuss thesisthrough theconference.However,due to the complexity of machine translation evaluation,there isno maturetestingplatform in the field of machine translation.The changes in the entire field of naturallanguageprocessing inrecent yearsare enormousand profound.In contrast,changesin thefield of machine translationare smaller.The mainreason isthat the fieldof machinetranslation coversalmost allthe mostdifficult problems in allareas ofnatural languageprocessing.Some unilateralor shallowbreakthroughs arenot enoughto haveafundamental impacton theoverall machine translation,only when theentirenaturallanguage fieldWhen allaspects ofresearch workhave achieved a comprehensivebreakthrough,it maylead toa breakthroughin machine translation.This is a processthatslowly accelerates,from qualitativechange toquantitative change,and graduallyaccumulates,which cannot be accomplishedovernight.
2.2Development of CAT
2.
2.1Past of CATComputer Aided Translation technologyis atemporary substituteproduced byMachine Translation development.Machine translation is atechnique in which acomputerconverts atext into another text.The disciplinearises fromIn the1950s,after halfa centuryofdevelopment,from arule-based approachtoa statistical andcase-based approach,thequality of translationhasgreatly improved.However,due to the complexityof naturallanguages,the readabilityof translationsis stillnot high.
2.
2.2Development of CATThrough thestudyof Computer Aided Translation,the followingdefects arefound:First,the alignmentof themodule is not ideal.Many CAT softwares quality of itsmodulesis uneven,requiring manualproofreading andincreased workload.The levelof automationofCAT softwaremust befurther improvedto ensureComputer Aided Translation quality.The secondis thelack of ability toextract technicalterms,especially theextraction ofmulti-word termsis abig problemfor theexisting technology.To solvethis problem,mutual informationand wordfrequency statisticsmethods can be used.Third,the retrievalabilityof QA retrieval toolsis notstrong.The currentQA searchtool on the marketcanonly retrievesimple errors.If theerror is a littlecomplicated,it cannot be retrieved.In thisregard,the use of collocationdictionaries and target languagecoipora forQAretrievaloftranslated contentis abetter approach.
2.
2.3Prospect ofCATAt present,the use ofCAT has become more and more extensive.Its convenientandaccurate translationhas madeit attractivein manyfields,especially engineeringand otherspecializedfields.It is only through the Internetthat thebalance betweentranslationefficiency and translation qualitycan be achieved.As analternative toMT,CAT has madegreat progressin theseyearsofdevelopment.The participationof humanshasmadetheerrors caused by the machine havebeen effectivelycorrected.CAThasunparalleled advantagesover humantranslation.Computers have the storage,retrieval,and computationalcapabilities thathumans cannever have.However,human beingsstill playa leadingrole inCAT.After all,human beingshave thewisdomand emotionsthat machinescannot possiblyhave.At the same time,humans havetotake correspondingresponsibility for the translation results.The developmentgoal ofCAT technologyistoconstantly improvethe automationlevelof translationtechnology.Although manis themain bodyof translation,it isnecessary to changethe computerintoasubject that is constantlychanging.Only inthisway canthe translation efficiency becontinuously improved.The currentautomation levelofCATsoftware is not high,and relatedtechnologies arerelatively simple.If wecancontinuously optimizethe systemand improvethe language model in the fieldof naturallanguageprocessing,then theimprovement of translation efficiencyis justaround thecorner.3-Different TechnicalPrinciples
3.1Technical principlesof MT
3.
1.1Rule-based Machine TranslationThe basicworking principleof Machine Translation isbased on the assumptionthatsentences withunlimited languagecan bederived from a limitedset ofrules.Machinetranslation methodsbased onthis assumptioncan befurther dividedinto threecategories:Direct Translation,Interlingual Approach,and TransferApproach.They allrequire the useof large-scale bilingualdictionaries,which requirethe use of source language derivationrules,language conversionrules,andtarget language generationrules.The differenceliesin thedepth of analysis of the language.For example,the literaltranslationmethodhardlyrequires linguisticanalysis,and theintermediate languagemethodandconversion methodrequiresome degreeof linguisticanalysis of the source language and the target language.
3.
1.2Statistics-based Machine TranslationBrown fromIBM firstused astatistical modelfor French-English machine translationin
1990.The basicidea isto considerthe machine translation problemas anoisy channelproblemand thenuse thechannel model to decodeit.The translation process isseen as adecoding process,which inturn becomesthe process of seekingoptimal translationresults.Machine translationbased onthis ideafocuses ondefining the most appropriatelanguageprobabilistic models and translationprobabilistic models,and thenestimating theprobabilityparameters oflanguage models and translation models.The parameterestimationof the languagemodelrequires a large amount of monolingualcorpus.Theparameter estimationof the translationmodelrequires alarge amountof parallelbilingualcorpus.The quality ofstatistical machine translationdepends largelyon theperformance oflanguagemodelsand translation models.In addition,good searchalgorithms arerequiredto find the optimaltranslation.In simpleterms,statisticalmachine translation firstestablishesastatisticalmodel,and thenuses the instances in the instancelibrary totrain thestatisticalmodelto obtain the required languagemodelsand translation modelsfortranslation.
3.
1.3Example-based Machine TranslationThe basicworking principleof Machine Translation isbased on the principleofanalogy.It matchesthe text segment mostsimilar to the source text segmentfrom theinstancelibrary,and extractstheinstancetextsegment.Corresponding target languagetranslation results,appropriate transformation,and ultimatelycome toa completetranslationresults.The coreidea ofcase-based machinetranslation wasfirst proposedbyMako Nagao.He proposedthat peopledonotmake deep-languageanalysisbut translatewhen translating simplesentences.First,the sourcesentence isdecomposed intoseveralsegments,and thenthese segmentsare translatedinto thetarget language.The translation ofeach segmentis obtainedby matchingwith theexample sentencestoobtainanalogy.Finally,these translatedsentences arecombined intoone longsentence.
3.2Technical principlesofCAT
3.
2.1Translation memoryTranslation memory is the core ofComputer AidedTranslation technology and animportant partof translation workstations.Bowker2002:93defines translation memoryas“a languagedatabase used to storeinformation sourcesand correspondingtranslations.^For thereference translationprovided by the system,if feasible,the translatorcan copyitcompletely and use itafter modification;if it is notfeasible,he cangive itup.Users useexisting text and translationsto buildone ormore translation memories.Inthe process of translation,the systemwill automaticallysearch the same or similartranslation resourcessuch assentences,paragraphs in the translation memory,and givereferencetranslations,so that the usercan avoidunnecessary duplicationof effortand onlyneed to focuson the translation ofnew content.Translation memoriesare constantlylearningand automaticallystoring newtranslations in the background,becoming more andmore“smart“and moreefficient.Almost alltranslation memorymanufacturers willtellusers:With TM,you neverhave to translate the same sentencetwice!A translation memory unitcontains the text sectionof thesource languageand itstranslation.These sectionscan betext blocks,chapters,sentences orsentences,or words.Individual wordsare treatedas specialterms andare generallynot in the domainoftranslation memoryeven if the translation memory stillcontains a single wordfor itstranslation memory unit.Research showsthat themultilingual filesthat manycompanieshave establishedonthe market usetranslation memorytechnology toassist.
3.
2.2Translation serverIn order tosolve the problem of multiple translators*participation inproject style andwording,in addition to extractingsegments with high terminologyand repetitionratebefore translation,an importantsolution isreal-time sharedmemory andterminologydatabases.Extraction technologyis notperfect,in manycases can not effectivelyidentifyall terms,especially multi-word terms.Therefore,in the translation process,oftenencountered after the term,then addedto thedatabase.To ensurethe consistencyof thetranslation,you canbuild The server allows translators to use real-time updatedtranslationmemories.Mainstream CATsoftware isprovided inboth stand-alone andserver versions.Unlikestand-alone translations,using the same databaseallows interpretersto sharetranslationmemory,as wellas translatorsand reviewers.Translations arefirst reviewedby reviewers.After enteringthe memorybank,it isusedbytranslators.This not only guaranteesthequality of translationmemory,but also accelerates theprogress of translation projects.
4.Advantages andDisadvantages of MT andCAT
4.1Advantages ofMTThe development of machinetranslation isso rapidthat there is asoil forsurvival inthe fieldof translation.Whats more,it hasmore advantagesthan humantranslation insomeaspects.
4.
1.1Low costMachine translation ismuch lessexpensive thanmanual translation.Because humantranslationrequires professional translators toperform the translation work,the costincludingtime cost,the expenditureto hireprofessionaltranslatorsis greater.Machinetranslation requireslittle manualintervention.Basically,the computerautomaticallyperforms the translation,which greatlyreduces thebudgetary costs.If youimport amanuscript into translationsoftware oran onlinetranslation system,you canget afreetranslation quickly.For translation of web pages,if therequirements arenothigh,thesource languagefiles can also be found onthe Internet.Then,compared withmanualtranslation,machinetranslationhas obviousadvantages,namely high speed andlow cost.This isvery attractivefor the translation ofmany corporatebusiness letters.
4.
1.2Ability toprocess complexinformation athigh speedIn the actualtranslation,it is inevitable thatsome urgentitems will be encountered.For example,a fewthousand words,but onlyhalf of the naturalday todealwith.Thedifficulty of this kindof workcannot bewell adaptedto anartificial translatorwho hasnoskill orexperience.Another exampleis the translation ofproduct descriptionsor brochures.There aresome wordsthat cantbefoundin thedictionary.This maybeaway ofsaying thatthe original authorhad anew combinationfor propagandaneeds.In general,translators arepowerless.In this case,ifthecompany can provide somehelp andprovide necessarybackground information,such asa corpus,the translatorcan usethe corpusand populartranslationsoftware,plus the necessary translationskills andknowledge reserves,it ispossible.Make translations of thesame qualityin lesstime.In practicalwork,workefficiency willincrease greatly.Today,the machinetranslation databaseis stillexpanding.With theever-increasingvariety of translation content,an ever-expanding databasecanprovidemore powerfulsupportfor alimited humanbrain.Moreover,as the translation of themachineis generallystable,this translationmethod can better ensurethat the translation will not bemisinterpretedby professionaltranslators in thefieldand becomea scrapmanuscriptinafixed setof manuscripts.Machine translation can increasethe quality of translations.
4.
1.3Supplement toHuman TranslationManypeople improperlythink thatif theyhold a dictionary inhand,they canuseforeign languagessmoothly,or they can justgraduate fromforeign languagemajors andtheycan directlyengage in translationwork.Neither engagedinthenecessary knowledgeof the translation industry,they haveno strictattitude of translation.This hascaused thevaryingquality of translators.Whether it is knowledge and skills,or workattitude,atpresent,thefieldof translationis stilllack ofsophisticated translators.In thebusiness world,translationworkis increasinglypursuing speed.There arefewerand fewertranslators who havetheattitude of“one isup andone isonthemoon.^^The translation of aliterary workis speededup andgiven toa numberof translators.Indoing so,the speedof translationis significantlyimproved,but there are oftencases ofinconsistenciesin convergenceand understanding.For example,“Amelia”is asimpleEnglish womanname thatis assignedto different translators.Some translatorsaretranslated intoEmilia1and othersare translatedinto Emelia”.From thetrust”point ofview,both usetransliteration methodsfor translation,so bothnames arein line with thestandard.However,for thetarget language audience inthiscase,Chinese readers,it maybequestioned that the latter is thesame personas theformer.Or thelatteris another person.We knowthat inthe nameofaperson,thesamesound and different glyphcan beregardedas anotherperson.If thewidely-used translationsof thisname areentered into the memorybank,the problem of inconsistencybefore andafterthe translation can be avoidedand thereaderwillnot have suchdoubts.
4.2Disadvantages ofMT
4.
2.1Language contextlimitations of machine translationInliterary translation,film and television translationis abranch withits owncharacteristics.Nowadays,moreandmore foreignfilm and television workshave beenintroducedinto China.In additionto the original texts,the translatedtexts of these worksmustalso takeinto accountthe characteristicsof the film and television worksthemselves.Because the textof film and television translationneeds tobe presentedto theaudience inaninstant,it will alsobelimited bythe screenspace.Machine translationcannot guaranteethat the translationisinlinewiththe lengthof timethe actorhas saidand thesize ofspacegiven bythe screen.Therefore,machinetranslationcannot satisfythe immediacyandspatial requirementsof film and television translation.In addition,the translation offilm and televisionto theaudience is not onlythemeaning ofa singlesentence,but moreimportantly,it conveysthe hugeculturalbackground information carried bythe lines.Different filmand television works havedifferentcreative backgrounds,reflecting differentcustoms.Such wordsas“How doIknow maybe saidby awell-mannered person,How doI knowAnd anilliterate drunkmaysay“Does theghost knowThe phrasemachinetranslationis likethis:“How doIknow Itcannot reflectthe charactersstate,mood oreducation levelinthetranslation.We knowthat thesame sentence,different peoplesay,will havedifferent results.Thecharacters inthe filmandtelevisionworks allhave distinctpersonality traits.Whentranslators aredoing translations,they mustshow theircharacter traits.This requirestranslatorswho havea solidbackgroundinbilingualism andwho arefamiliar withthe basicqualitiesof thetwo countriesto tryto figureout thecharacter,make correctjudgmentsbased onthe circumstancesbefore andafter,and thenselect appropriatevocabulary fortranslation.This machinetranslationiscurrently impossibleto achieve,because themachinecannotdistinguish thecharacter ofothers,can onlytranslate fromliterally.Although thesource languagelooks thesame,but due to theabove-mentioned manyreasons,in order to completelytransmit the informationcarriedbythesourcelanguage,butalso tomanually process thetranslation.
4.
2.2Functional inequivalenceInliterary works,thelanguageusually hasshade ofmeaning andrhetoric.Film andtelevisionworks areno exception.Where the rhyme isinthe original speech,then thecorrespondingtranslation shouldalso payattentionto therhymeof thetext.To achievereciprocity,the translatorneeds tohaveasolid bilingualbackground.Voice-over,which iswhatwe oftensay.If it isonlytranslated literally,it maybe difficultfor thetarget audienceto understand.In addition,filmandtelevisionworksoften havea commercialnature,andits audiencemust beas extensiveas possible.This requiresthat thetranslationof the workmustbe easytounderstand.Machine translationcan only be translatedliterally,and it is notpossibleto simplifycomplex languages.Translators cando thiswhentranslating,but tograspthe degreeof goodness,itisnecessaryto“domesticate sothatthetarget audiencecaneasily obtain thesameviewing experiencewhile retainingthenecessary“alienation”.The target languageaudiencecanalsoexperience exoticismunder thepremise ofbeingable toread thevideo.The excellenttranslationoffilmandtelevision requiresa veryhigh leveloftranslators,notonlythe abilityto usethelanguage,but moreimportantly,the translatorspersonalexperience andcomprehension of the work.At thispoint,machinetranslationbasicallydoes nothing.The biggestdrawback of machinetranslationisthatit hasno independentfeelings.If allthe filmandtelevisionworks arehanded overto themachine for translation,it willnotbe abletotransfer alltheinformationcovered inthefilmandtelevisionworks tothetargetlanguageaudience.What wewill gainisalack ofemotional works.Functional equivalencymeans thatthetranslation should make the responseof theaudience of thetranslation filmbasically thesame asthe responseof theaudienceof theoriginal filmandtelevisionworks tothe original text”,such ascomedy translation.Comedyneeds toachieve theeffect ofcheering theaudience.If thetranslationofcomedy isboring,then therewill bea situationwhere thesourcelanguageaudience laughsforward andbackwards,but thetargetlanguageaudience cannotobtainthesame feelingand thecloudlesssituation.From thispoint ofview,machinetranslationdoes notmeet therequirementsof functionalequivalence.AcknowledgementI wouldlike toexpress mysincere appreciationto mysupervisor,Jessie Li,for heracademicguidance andpersonal supportthrough everyphase ofthis dissertation,and alsoforher encouragingmy accomplishmentof thewhole research.Without hergreat patienceandhelpful suggestions,the completionofthis dissertation wouldnothavebeen possibleandsmooth.I wouldlike toextend mygreat gratitudetotheauthors whoprovide numerousarticlesfor reference,and letme standontheshoulders ofgiants.Without theenlightenment andhelp,I cannotaccomplish mydissertation allon myown.I amalso indebtedtotheleadership of the Schoolof ForeignLanguage,JilinUniversity ofFinance andEconomics,for theirsupport andunderstanding throughouttheyears ofmy studyingand workingon campus.Last butnot least,I alsowant toexpress mysincere appreciationto mydearclassmates for their valuablesuggestions andhelp.
4.3Advantages ofCATWith thecontinuous increasein translationrequirements,traditional manualtranslationcannot keeppace withthe times,and the use ofcomputer-assisted manualtranslationhas its own advantages.
4.
3.1Improvement oftranslation efficiencyThecoreofComputer Assisted Translation technologyis translationmemorytechnology.When translatorsare constantlytranslating,Computer AssistedTranslationsoftware storesthe original textandtranslationina languagedatabase,which isatranslation memory.In the processoftranslation,whenthesame orsimilar sentencereappears,the systemwill automaticallyprovide theoriginal textandtranslationto thetranslator,and thetranslator will beableto maketrade-offs asneeded.Or inthe nexttranslationof thesame type of document,theoriginaltranslationmemorywillbeadded totheComputer AidedTranslation software,the systemwill automaticallypre-translate,andthe translatorstask istotranslatethe remainingpart.Therefore,using Computer AssistedTranslation techniques,translators onlyneed totranslate newcontent,which greatlyimprovesthe efficiency oftranslation.At thesame time,ComputerAssistedTranslation technology can keep the translationinthesameformat astheoriginal.According tothe traditionalmodel,when thereare chartsormarks intheoriginal text,the translatorsneed toprocessthecharts ormarks againinWord orother software.Once a very complicatedchart isencountered,the interpreterwillspend a lot oftime onthe drawings,and thesetimes arenon-productive times.However,using ComputerAssistedTranslation technology,translators donot need topayattention totheformat ofthetext,because thetechnologycankeep thetranslation formatconsistentwith theoriginal format,thereby reducingalotof non-production timeand improvingthetranslation efficiency.
4.
3.2Cooperation ofseveral translatorsIn order tomaintain thetranslation styleand wordconsistency,the traditionalmanualtranslation isdone bya singleinterpreter.Nowadays,duetothe largeamount oftranslationsand thetight translation time,asinglemember whocompletes atranslationproject alonecannot meetthemarketdemand.However,different translatorsuse differentlanguagestyles and use differentwords.It isvery difficult to keepthetranslationconsistent.Computer AidedTranslation provides a solutionto thisdilemma.In Computer AidedTranslation software,the translators can communicatewith eachother and can seetheglossaries andtranslationsofother translators,thus ensuringthatthe translators1translations areconsistent instyleandwording.As aresult,Computer Aided Translationtechnology hasenabled theconversion oftraditional single-person translationsintotranslators collaborativetranslations.
4.
3.3Guarantee ofconsistent translationFor thetranslationofadocument,in additiontothetranslator,there isa review.Whenmany peoplecooperate intranslation,because ofdifferenttranslators,different styles,anddifferent wordchoices,itisdifficult tomaintain unityoftranslation,and itwillalsobringenormous workloadtothe review.In additionto translationmemory,Computer AidedTranslation technology alsohas an important component,terminology management.Ifwords appearingintranslationneedtobe reused,theycan be savedas terms.Forthesameproject,therequiredterminology canbe unifiedin advanceand thensent tomultipletranslators sothat boththe choiceof termsand the terminology canbe consistentand thedisambiguationof wordsis eliminated.Through thetranslationmemory,thetranslatorscanalso seethesameorsimilartexts thathavebeentranslated,which cankeepthe translationstyle consistent.Therefore,Computer AidedTranslation technologycan ensuretheconsistency oftranslation,and alsoreduce the workload forthereview.
4.4Disadvantages ofCATAlthough Computer AidedTranslation technology hasso manyadvantages,due tothecomplexityoflanguages,therearestill deficienciesin Computer AidedTranslationtechnology.
4.
4.1Hardening translationThe software inComputer AidedTranslationtechnologyseparates theoriginal textbysentences.In thetranslation process,translatorscanonly performone-to-one translationsaccordingto separatesentences.Because thetranslationcanonly beperformed directlyaccordingtotheoriginaltextsequence,and theoriginaltextcannotbegreatly adjusted,thetranslators subjectivecreativity islimited.Moreover,the translationstranslated bythistranslation methodmakethearticles lackintegrity and thetranslationsappear tobestiffened.
4.
4.2Limited applicationdocuments with high content repetition ratesOne oftheadvantages ofComputer AidedTranslationtechnologyistheuse oftranslationmemories andterminology databasesto increasethe efficiency oftranslation.However,this techniqueismoreapplicable to documents withless sentencechanges andhighcontent repetition rates,such aslaws andtechniques inthetranslationof non-literarytexts,manuals,etc.Computer AidedTranslationtechnologyisnotsuitable forthetranslation ofliterary texts,because literarytexts havemany variationsin sentencepatterns,rich rhetoric,and closelyrelated contexts,making itdifficult topre-translate orfrequentlyusing translationmemories andterminology databases.
4.
4.3Low obscurematching recall rateWhen searching,you needto searchthe memorybank tofind asuitable translation.There aretwo kindsof resultsforthetranslation machinesearch,namely exactmatch andfuzzy match.An exactmatch returns100%ofthenew textand thesame memorysegment,and thefuzzymatchreturns asimilar segment.In areal translation project,most ofthe highrepetitionratesare technicaltexts such asdescriptions.But mostprojects arehard tofind exactlythesamecontent.Moreover,thesoftware currentlyused inComputer AidedTranslationtechnology has lowrecall offuzzymatching,such asTrados fuzzy matching recallrate is70%,and sentenceswith similaritylessthan70%cannot automaticallyreturn.However,whether itis Englishor Chinese,itssentence patternsare complexto change.The lowfuzzymatching recallratemakessentences thatcanbeautomatically returnedlimited,so fortexts with low repetitionrates,it stillrequires manual translationofalargeamountofcontent.Although Computer AidedTranslationtechnology hasachievedalot,ComputerAided Translationtechnologyhastobedeveloped toamoreintelligent level,and itwilltake along timeto resolveitsownshortcomings.5-Different Applications
5.1Applications ofMT
5.
1.1Translating machineFromtravel abroad,to internationalcultural exchanges,to foreigntrade,languagebarriers area naturalpain point.Many currenttranslation productscombine machinetranslation with OCRtechnology andspeech recognitiontechnology.By translatingforeignlanguage signs,menus,manuals,etc.throughthe camera,itis also possibletoengage inconversational translationsin conjunctionwith speechtechnology toachieveunobstructed communicationin different languages.
5.
1.2Simultaneous-interpretation technologySimultaneousinterpretation iswidely used in internationalconferences andothermultilingual communicationscenes,but artificialinterpretation islimited byfactors suchasmemory,listening andspeaking speed,and highcost.The machine-simultaneoustransmission technologyof Sogouwas launchedlast year.Gradually beganto appearin theconferencescene,the speakersspeech isconverted intotext inreal time,simultaneoustranslation isperformed,and thetranslationresultsare displayedwithlowdelay,hoping toreplacethe artificialinterpretation andachieve low-cost andeffective communicationamongdifferentlanguages.
5.
1.3Cross-language retrievalChineseinformation onlyaccounts for10%ofthe worlds information,andcross-language retrievalneeds increaseyear byyear.Sogous overseassearch launchedearlierthis yearcombines machinetranslation andinformation retrievaltechniques.Whether usersinput Chineseor English,the systemwill Selectthe usersdesired searchresultsfromalargenumberof high-quality Englishwebpages,andusethe worldsleadingmachine translationto automaticallytranslate them,providing userswith Englishtext,Chinese translations,and bilingualChinese andEnglish searchresults.
5.2Applications ofCATIn orderto copewiththe problemsinmodern translation,CAT technologycame intobeing.The CATsoftware withtranslationmemoryasthecore,first solvedthe problemoftranslation multiplexing,shortened thetranslationtime;established theterminology base,unified terminology;useofquality assuranceQuality Assurance,QA Toolsto improvetheefficiencyofreviewing andupdating,and to update andmaintain memoryatthesametime;Translation ManagementSystem TMSto managethetranslationprocess andcoordinateproject participants;TMS softwarefor server,toupdatetheterm bank inrealtime bybuilding atranslation serverAnd memorybank,convenient multi-interpreterproject a unified styleanduseof words,but alsoallows auditorsand translationsynchronization.
5.
2.1Memory technologyTranslation Memory ExtensibleMarkup LanguageXML,which alignsthe originaltextandtranslationof existingtranslation materials,and addstags tothetextto achieveretrievaland extractionof unstructureddocuments.In orderto facilitatethe exchangeoftranslation memoryin differentCAT tools,aunifiedTranslationMemoryExchangeFormat TMXwas developedbasedonthe XMLby Languageand LocalizationStandardsAssociation LISA.In additionto TMX,LISA hasalso developeda TBXformat forterminologyexchange.The term bank isalso anapplication oftranslationmemory.Unlike thetranslationmemory,the contentoftheterm bankismainlyused asa reference,whenopening anew segment,CAT Thetool automaticallyretrieves thetermbankand returnstherelevant terminology.The translatorthen decideswhether touse itinthetranslation.The contentofthetermbankcanbeimported fromadictionary,or itcanbeextracted byaCAT toolbefore translation,aunifiedmanualtranslation,ortranslationManually addedduringtheprocess.The applicationoftranslationmemory reducesthe repetitivework inthe workof thetranslatorand improvestheworkefficiency.However,thetranslationmemory alsohaslimitations.First ofall,it onlyplays agreater role inthetranslationwitha higherrepetitionrate,and manytranslations haveno similarity.Second,fuzzy searchtechnology needsto bedeveloped.
5.
2.2Translation serverInordertosolve theproblemofmultiple translators1participation inproject styleandwording,besides extractingsegments withhigh terminologyand repetitionrate beforetranslation,animportantsolution isreal-time sharedmemory andterminology databases.Extraction technologyisnotperfect,in manycases cannot effectivelyidentify allterms,especially multi-word terms.Therefore,inthetranslationprocess,often encounteredafterthe terms,then addedtothedatabase.To ensurethe consistencyofthetranslation,you canbuildTheserverallows translatorstousereal-time updatedtranslationmemories.Mainstream CATsoftwareisprovided inboth stand-alone andserver versions.Unlike stand-alone translations,theuseofthesame databaseallowstranslatorsandreviewers towork inadditiontosharing translationmemory betweentranslators.Translations arefirst reviewedby reviewers.After enteringthe memorybank,itisused bytranslators.This notonly guaranteesthe qualityoftranslationmemory,but alsoacceleratesthe progressofthetranslationproject.
5.
2.3ReviewIn traditionaltranslation,thereisa post-reviewer.In smallertranslation projects,thereview workis generallycompleted bytranslators.In moderntranslation,the scaleof theprojectbecomes largerand larger,and the reviser becomesan independentposition.Theworkload ofthereviserAs muchas aninterpreter,a reviseris oftenresponsible formultipletranslators.Therefore,reviewers alsoneedtouse computer-assisted workto improveworkefficiency.Quality assurancetools aresoftware thatassists reviewersin theirwork.Such asQA tools.The QAtool canhelp thereviewer detectthe basicerrors inthetranslation,and theseerrors areoften importanterrors.The generalQA toolsinclude thedefaulterror typessuchasmissing translations,terminology errors,inconsistenttranslations ofthesamesourcetext,and formaterrors.Some QAtools allowthe user todefine thetypeoferror andthe systemautomatically findsit.Taking XBench asanexample,the softwareprovidesarelatively simplecustom mode.The usercan definetheerror byinputting theoriginaltextandtranslationthat hewants tofindtheerror.Theretrieval methodis clumsyand canonlybe used ifthe userpredicts possibleerrors.QADistiller providesmore flexiblecustom error types.The softwareallows theusertouseregular expressionsto defineerrortypes.This methodallows usersto Defininga classoferrors.Of course,this kindof operationis suitablefor userswhohavea certainbackgroundin computerknowledgeandunderstand thesyntax ofregular expressions.ConclusionWith thedevelopment ofthe globaleconomy andthe increase of exchangesbetweencountries,the demandfor language translation willincrease further.Under suchabackground,how tohandle languageconversion moreefficiently andwithhighquality hasbecomea commonconcern ofmany organizationsand companies.Although thereareproblems withmachinetranslationinoneway oranother,its translationsometimes failstoachieve thefirst standardof faithfulness,expressiveness,and elegance,but itseaseofuseprovides peoplewith greatconvenience.It hasfunctions suchas webpage translation,Chinese-language software,internal codeconversion,collation translation,and soon.Ithas playedaveryimportant rolein helpingusers performword processing,web browsing,reading orcomposing e-mails.On theother hand,Computer AidedTranslation ofadvancedmemory technologies,convenient human-computer interactionmechanisms,andopen vocabularies,memory banksand grammarlibraries provideprofessional translatorswithan efficient,high-quality controltranslation platform.Greatly improvethe efficiencyandqualityoftheir translation.With thedevelopmentofscience and technology,the technologiesusedin machinetranslation andcomputer-assisted translationare constantlyupdated andare constantlyevolving,and willplay anincreasingly importantroleinthe futureoftranslation.ReferencesBaker,M.
1999.The roleof corporain investigationthe linguisticbehavior ofprofessionaltranslators.International Journalof CorpusLinguistics,42:281-
298.Bowker,L.
2002.Computer-aided TranslationTechnology:A PracticalIntroduction.Ottawa:University ofOttawa Press.[]Brown PF.
1990.A StatisticalApproach to MachineTranslationJ.ComputationalLinguistics,162:79-
85.Hutchins,J.Somers,H.
1992.An InstructiontoMachineTranslation.London/SanDiego:Academic Press.Nagao,M.
22.
1.1Initial stage1946-
22.
1.2The twistsand turns1954-
32.
1.3Recovery phase1975-
32.
62.
62.
62.
73.
73.
73.
83.
83.
104.
104.
104.
124.
124.
144.
144.
144.
154.
154.
164.
175.
175.
175.
185.
185.
195.
1.Literature ReviewMachineTranslation,MT forshort,isasub-fields ofcomputational linguisticsthatinvestigates theuseofsoftwaretodirectly translatearticles orlectures fromsourcelanguage totargetlanguage;While,ComputerAidedTranslation,sometimes referredto bytheabbreviation CAT,isaform oflanguagetranslationinwhichhuman translatorsmakeuse ofcomputer softwareto advancethetranslationprocess.In thepaper ofCarberry S:In Proceedingsofthe23rd AnnualMeeting oftheAssociation forComputational Linguistics,hecametoaconclusion thatMachinetranslation is essentially acomprehensive research field.It isrooted inthe intersectionofscience andengineering,basic researchand practicaldevelopment,computer science,linguistics,artificial intelligence,and softwareengineering.It isalsoatheoreticallinguistics andcomputational language.The mostnatural applicationof learninganddescriptive linguistics.Machine translationisamultidisciplinary productthat combinestruescience withtechnology.The studyofmachinetranslation willcertainly promote therapid developmentof thesedisciplines.Inthearticle《机器翻译概述》written byLv Xueqiang,Machine translationtechnologyhas far-reaching significancein acceleratingand expandinginformationdissemination aroundtheworld.The fieldofmachinetranslation alsoprovides aconvenientand idealtest environmentforthequality assessmentof knowledgebases,linguistic theories,and linguisticdescription studies,andcanbeusedasadesign toachievelarge scaleThesoftwareengineering experimentsite ofthe softwaresystem.Thecomponents ofthe comprehension-based machinetranslationsystemcanbedirectlyapplied toother applicationsinthefields ofcomputationallinguisticsand artificialintelligence.Machine translationisessentiallyan integratedresearchfield,and itis rootedinscience andengineering.The intersectionofmultiplefields suchas basicresearch andpracticaldevelopment,computer science,linguistics,artificial intelligence,and softwareengineeringisalsothemostnatural applicationof theoreticallinguistics,computationallinguistics,and descriptivelinguistics.Machine translationisareal Thecombination ofscienceandtechnologywith multidisciplinaryproducts.The studyofmachinetranslationwill certainlypromotetherapid developmentof thesedisciplines.But someconclusions whichgot bysome scholarswere outdated.Paper aboutthecomparison areinadequate,or notdetailed.So Istart myarticle aboutthe comparisonstudyon MachineTranslationandComputerAidedTranslation forthe detailedstudy andadequateinformation.
2.Development ofMT andCAT
2.1Development ofMT
2.
1.1Initial stage1946-1954In fact,as earlyasthefirst digitalelectronic computerestablished in1946,peoplebegan to think abouttheproblemof usingcomputers insteadof peopletodotranslationwork.Even beforethis,Turing hasalready beguntothinkabout whethercomputers canthink.Oneofthe landmarkevents ofthis stagewas the1949information pioneer:WarrenWeaver.The memorandumon machinetranslation waspublished,andthecalculability ofmachinetranslation wasput forward.The secondcase wasthat,in1954,GeorgetownUniversity underthe cooperationof IBMconducted anEnglish-Russian translationexperimentand beganan attemptontranslationautomation.At this stage,in general,theconcept ofmachinetranslationhas beenformed inpeoples minds,and peoplenaturallythink oftheuseof grammaticalrules anddictionaries fortranslation purposes.People areoptimisticthat byexpanding vocabularyand languageregulations,theproblemof machinetranslationwillbebetter resolvedinthenear future.So aftersome time,countries havestronglysupported machinetranslation projectsand formeda machinetranslation study.
2.
1.2Twists andturns1954-1975In November1966,ALPAC UnitedStates LanguageAutomatic ProcessingAdvisoryCommittee evaluatedvarious translationsystems fromthe perspectivesof speed,quality,and costofmachinetranslationandthe needfor machinetranslation atthat time,and publishedthe famousAPLAC report,which gavemachinetranslationresearchworkahead-on blow.According tothe report,thetranslationqualityofmachinetranslationisobviously lowerthan thatof humantranslation.Machine translationencounters asemanticbarrier”thatisdifficulttoovercome.The reportcompletely negatesthefeasibility ofmachinetranslationand recommendsthat investmentinmachinetranslationbe stopped.Although theconclusions ofthe reportwere toohasty andarbitrary,theresearch atthisstagedid notsolve manyofthemost importantissues anddid notprovidean in-depth analysisofthelanguage.Since then,intheworld,machine translationshaveseen anunprecedented depression.
2.
1.3Recovering phase1975-1989With theadvent ofmicroprocessors,computer capabilitieshave grownby leapsandbounds.Machinetranslationhas beenreintroduced.Because ofits hugecommercialinterests andthe participationof largecompanies,it resultsintherecovery anddevelopmentofmachinetranslation.The studyofgrammarand。
个人认证
优秀文档
获得点赞 0