Mona Baker & Gabriela Saldanha. Routledge Encyclopedia of Translation Studies (2nd edition)

Подождите немного. Документ загружается.

Page168

specialpurposesystems(DIPLOMAT),e.g.fortranslationofSerboCroatian

inmilitaryoperations(Hutchins2001b).

AnumberofsystemsthatintegrateCATtechnologiesintoMTweredeveloped

inthepasttwodecades.TheInstituteofComputationalLinguistics(ICL)of

PekingUniversity(PKU/ICL),theInstituteofComputingTechnologyofthe

ChineseAcademyofSciences,andtheStateKeyLaboratoryofIntelligent

TechnologyandSystemofTsinghuaUniversityatBeijingjointlydevelopeda

Chinese–EnglishMTsystem,primarilyorientedtowardsjournalistictranslation,

whichincorporatescorpusbasedtranslationmemory(TM)technologiesina

multipleenginearchitecture(Baietal.2002:124).Fujitsu’ssoftwarepackage

ATLAS,whichtranslatesfromJapanesetoEnglishandEnglishtoJapanese,

alsousesbothMTandTMtechnologies(ATLASV13).

SinceEurotra,aresearchprojectfundedbytheCommissionoftheEuropean

Communities,failedtoproducetheworkingsystemthattheCommissionhoped

wouldreplacethecommerciallydevelopedSYSTRANsystemsithadbeen

using,researchfundedbytheEuropeanUnionhasfocusedmoregenerallyon

projectswithinthebroadfieldoflanguageengineering.Manyofthese

multilingualprojectsinvolvetranslationofsomekind,usuallywithinarestricted

subjectfieldandoftenincontrolledconditions(Hutchins2005b).However,MT

isusuallynotusedalonebutalongwithcomputeraidedtranslationandother

(multilingual)informationprocessingtechnologies,suchasconcordancersand

terminologymanagementtools,whichareusuallycompatiblewiththeword

processingsystemsprofessionaltranslatorscustomarilyuse,andareintegrated

in‘translatorworkstations’.

Promisingresearchdirections

AmongthemanyresearchdirectionsexploredinMTinthepastdecadeorso,

twostandoutasparticularlypromising:knowledgebasedmachinetranslation

(KBMT)andstatisticalmachinetranslation(SMT).

Knowledgebasedmachinetranslation(KBMT) 

Artificialintelligenceresearchdiscoveredthataddingheuristics(rulesofthumb)

enabledcomputerprogramstotackleproblemsthatwereotherwisedifficultto

solve.Thediscoveryinspiredeffortstobuildknowledgebasedsystemstohelp

solvetraditionalproblemsinMT.Aknowledgebasedsystemisbasedonthe

methodsandtechniquesofAIandisprogrammedtoimitatehumanproblem

solvingbymeansofartificialintelligenceandreferencetoaknowledgebase

(KB),i.e.adatabaseofknowledgeprovidingthemeansforthecomputerized

collection,organizationandretrievalofknowledgeonaparticularsubject.The

corecomponentsofaKBMTaretheKBandtheinferencemechanisms.

Developingrelevantknowledgebasesorresources(includingdictionariesof

grammaticalandsemanticrules,specializeddictionariesorglossaries,term

banks,translationmemories,alignedparallelcorpora,worldmodels,etc.)is

expectedtoimprovetheaccuracyofMToutput.Adictionarythatcontains

usefulinformationonwordsegmentation,forexample,candramaticallyreduce

thecomplexityusuallyassociatedwiththeuseofaparserinsegmentation(Wu

andJiang1998:1).Characterizedbyatightintegrationofautomatictranslation

technologieswiththeexpertiseandexperienceofhighlyskilled

linguists/translators,KBMTisviewedasabridgebetweenthetwoextremesof

humanonlyhighqualitytranslationandmachineonlylowqualitytranslation,

andgenerallyyieldshigherqualityoutput.

ThebestknownKBMTprojectisKANT,foundedin1989attheCenterfor

MachineTranslationatCarnegieMellonUniversity.Oneexampleofare

designed,objectorientedC++implementationofKANTtechnologyforMTis

KANTOO,whichfeaturesatool(‘theKnowledgeMaintenanceTool’)for

knowledgesourcedevelopment.Intendedprimarilyforthedeveloperandend

usermaintainer,thetoolprovidesaninterfaceforstructurededitingofgrammar

rulesanddomainknowledgeandcanbeusedbycustomerswhowishto

continuecustomizingthegrammaranddomainknowledgeafterthedeliveryofa

finishedsystem.

TheInstituteofComputationalLinguisticsatPekingUniversity(PKU/ICL)has

plansto

Page169

buildanintegrated,comprehensiveLanguageKnowledgeBasetosupport

Chineseinformationprocessingtasks(includingChinese–Englishtranslation).

TheKnowledgeBasewillconsolidateanumberofimportantlanguagedata

resourcestheInstitutehasdevelopedinthepasttwodecades,whichinclude,

amongotherthings,theGrammaticalKnowledgebaseofContemporary

Chinese(GKB),thePOSTaggedCorpusofContemporaryChinese,the

SemanticKnowledgebaseofContemporaryChinese(SKCC),theChinese

ConceptDictionary(CCD),abilingualparallelcorpus,andamultidisciplinary

termbank(Yuetal.2004).KBMTtechnologieshavealsobeenappliedin

commercialsystemssuchasSDLInternational’sSDLKbTSystemand

CaterpillarInc.’sCaterpillarCorporateTranslations.

Statisticalmachinetranslation(SMT)

Statisticalanalysisofhuge,alignedbilingualcorporaallowsfortheautomatic

constructionofMTsystemsbyextractinglexicalandsyntactictranslation

equivalentsfromsuchcorporaonastatisticalprobabilitybasis.Forlanguage

pairssuchasChinese–EnglishorArabic–English,statisticalsystemsarealready

thebestMTsystemscurrentlyavailable(SMTGroupattheUniversityof

Edinburgh2006).

ThestatisticalapproachtoMThasbeendevelopedbyteamsatIBM,Johns

HopkinsUniversity,UniversityofPennsylvania,andtheInformationSciences

InstituteoftheUniversityofSouthernCalifornia(USC/ISI),amongother

partners.KnightandMarcuatUSC/ISI,forexample,devotedtwentyperson

yearstothedevelopmentofSMTsystems.ThekeytotheirSMTsoftwareis

thetranslationdictionaries,patternsandrules(knownas‘translation

parameters’)thattheprogramdevelopsandranksprobabilisticallyonthebasis

ofpreviouslytranslateddocuments.KnightandMarcufoundedacompany

calledLanguageWeaverwhichsellssystemsforArabic,Chinese,French,

German,Persian,RomanianandSpanishtranslationtoandfromEnglish(Knight

2005;Hutchins2006).

GoogletoohasreportedpromisingresultsobtainedusingitsproprietarySMT

engineanditsmassivetextdatabases.ItcurrentlyusesSYSTRANformostof

thelanguagepairsithandles,butisworkingonastatisticaltranslationmethodto

implementinmostofitsonline‘GoogleTranslate’servicesinthefuture.

AlthoughthestatisticaltranslationmethodisnowusedonlyinArabic–English,

Chinese–English,Japanese–EnglishandKorean–Englishtranslation,more

languagepairswillsoonbemigratedfromtheSYSTRANenginetothe

statisticsbasedGoogleengine.In2007,Googleimprovedthisengine’s

translationcapabilitiesbyinputtingapproximately200billionwordsfromUnited

Nationsmaterialstotraintheirsystem.Theaccuracyoftranslationsprovidedby

Googlehassinceimproveddramatically(Hutchins2006).

Seealso:

COMPUTERAIDEDTRANSLATION;CORPORA;LOCALIZATION.

Furtherreading

HutchinsandSomers1992;Arnoldetal.1995;Trujillo1999;Hutchins2001b;

Liu2002;Quah2006.

KEPING

Minority

Therelationshipbetweentranslationandminoritylanguageshasbeenarelatively

neglectedtopicformuchoftheexistenceoftranslationstudies.Translation

theoryanthologiesrarelyincludedcontributionsfromminoritylanguage

perspectives,andlittleornoallowancewasmadeforthefactthatattitudes

towardstranslationmightsignificantlyalterdependingonwhetherthesourceor

targetlanguagewasinamajorityorminorityposition.Althoughanumberofthe

significanttheoristsintranslationstudiesinthe1970sand1980s,forexample,

camefromsmallercountriessuchasBelgium,IsraelandtheNetherlands,this

didnottranslateintoaspecificconcernwiththepositionofminoritylanguages.

Itwasintheareasofanthropology,areastudies,literaryandculturalstudies–

ratherthanintranslationstudiesperse –thatquestionsbegantobeasked

abouttherelationshipbetweentheroleoftranslationforcommunitiesina

subordinate