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

Подождите немного. Документ загружается.

Page60

whereeachwordinarunningtextisassignedtoawordcategory(e.g.‘noun’

or‘verb’),aswellassyntacticparsingandsemanticannotation.Anumberof

sources(Kenny2001;Meyer2002;Olohan2004)provideguidanceonthe

creationandannotationofcorpora.Kilgarriffetal.(2006)isparticularlyuseful

forthoseinterestedinincludingwebdataintheircorpora.

Thelevelofmarkupthatacorpusissubjectedtowillhaveimplicationsforthe

kindofelectronicprocessingthecorpuscanundergo.Rawcorpora–thatis,

untagged,unparsedcorpora–canbetreatedassequencesofcharacters

delimitedbyspaces(inlanguageslikeEnglish);inotherwords,assequencesof

runningororthographicwords.Thenumberofdifferentorthographicwords

(types)inacorpuscanbeeasilycomputedandcomparedwiththetotalnumber

ofrunningwords(tokens),toyieldthetype–tokenratio,aprimitivemeasureof

thelexicalvarietyinatext(seeKenny2001andDalleretal.2003onthe

limitationsoftype–tokenratios).Valuesforaveragesentenceandparagraph

lengthscanalsobefairlyeasilycomputedinarawcorpus(usingsentence

endingpunctuationmarksandparagraphsymbols).Anothermeasure,lexical

density,givesthepercentageofrunningwordsinacorpusmadeupbylexical

(vocabulary),asopposedtogrammaticalwords.Putsimply,alowlexical

densityindicateshighlevelsofredundancyandthuspredictabilityinatext

(Stubbs1996:73),perhapsmakingiteasiertoprocessthanalexicallymore

densetext.Aslongastheanalystcansupplyafinitelistofgrammaticalwordsto

beexcludedinthiscalculation,lexicaldensitiesarealsofairlyeasilycomputed.

Allthesemeasureshavethesignalmeritofbeingapplicabletoeveryverbaltext,

buttheirinterpretationisnotstraightforward,andtheyshouldbeusedwithcare.

Othertypesofprocessingallowmorequalitativeanalysisofcorpusdata.The

generationofwordlistsandclusters,forexample,allowstheanalysttofocuson

particularwordsorrecurringgroupsofwords.Afurthertypeofprocessing

outputsaKWIC(keywordincontext)concordanceforaninputword,

revealingthecontextsinwhichitoccurs.Recurringpatternsmaybediscerned

acrosscontexts,pointingtotheexistenceofstatisticallysignificantcollocatesof

theinputword.Inordertodiscoverregularitiesinthebehaviourofawordform

occurringinacertainpartofspeech(seeOlohan2004:70–71),however,a

taggedcorpusisusuallyrequired.Concordancingandbasicstatistical

processingofrawcorporaarediscussedinSinclair(1991).Moretranslation

orienteddiscussionsareavailableinKenny(2001)andOlohan(2004).

Translationorientedcorpustypology

Severalscholarshaveproposedcorpustypologiesthatareofparticular

relevancetotranslationstudies(see,especially,Laviosa1997,2002).Atahigh

levelofabstraction,corporacanbedividedintothosethatcontaintextsina

singlelanguage–monolingualcorpora–andthosethatcontaintextsintwoor

morelanguages–bilingualormultilingualcorpora,although,forthesakeof

economy,bilingualcorporaaresometimessubsumedunder‘multilingual’

corpora(AltenbergandGranger2002:7).Anotherbroadcharacterization

dependsonwhetherthetextswereoriginallywritteninthelanguageinquestion,

orwhethertheyweretranslatedintothatlanguage.Byfarthebestknown

corporaoutsideoftranslationstudiesarelargemonolingualreferencecorpora

liketheBritishNationalCorpus(Burnard2007),whichcontains100million

wordsoriginallyutteredorwritteninBritishEnglish.Laviosa(2002:37)calls

suchcorporanontranslational.TheTranslationalEnglishCorpus(Olohan

2004:59–60),ontheotherhand,isperhapsthebestknownmonolingual

translationalcorpus.Itcontinuestobedevelopedunderthestewardshipof

MonaBakerattheUniversityofManchester,andcurrentlycontainsroughlyten

millionwordsoftexttranslatedintoEnglishfromavarietyofsourcelanguages.

Corporamayalsobecharacterizedbytherelationshipthatholdsbetweentheir

subcorpora,wheretheseexist.Thus,amonolingualcorpusmayconsistoftwo

subcorpora;onetranslational,theothernontranslational.Ifthetwosetsoftexts

coverthesamegenre(s)inroughlythesameproportions,werepublishedinthe

sametimeperiod,coverthesamedomains,etc.,thenwecanspeakofa

monolingualcomparablecorpus.Monolingualcomparablecorporaallow

systematicinvestigationsofhowtranslatedtextdiffersfromnontranslatedtext

in

Page61

thesamelanguage,andthusareavitalresourceinresearchthatseekstoisolate

characteristicfeaturesoftranslation(seebelow).Wellknownmonolingual

comparablecorporaincludeLaviosa’sEnglishComparableCorpus(1998a,

1998b)andtheCorpusofTranslatedFinnish(Mauranen2004).Likewise,the

subcorporainabilingualcorpusmayberelatedthroughsharedvaluesfor

attributessuchasgenre,dateandplaceofpublication,domain,etc.,andthus

combinetoformabilingualcomparablecorpus.TheNewCorpusforIreland

(Kilgarriffetal.2006),designedinthefirstinstanceasaresourceforEnglish–

Irish(Gaelic)lexicography,isonesuchcorpus.Bilingual(ormultilingual)

comparablecorporaaresometimesusedasadatasourceincontrastive

linguistics,andarevaluedpreciselybecausetheyarefreefrom‘various

translationeffects’(AltenbergandGranger2002:8).Theyarenotwithout

problems,however:aswithmonolingualcomparablecorpora,itcanbedifficult

toensurecomparabilitybetweenthesubcorpora(seeBernardiniandZanettin

2004),andsearchingfor‘crosslinguisticequivalents’(AltenbergandGranger

2002:9)isnotstraightforward.Baker(1995:233)hasalsoexpressed

reservationsabouttheirusefulnessintheoreticaltranslationstudies,claimingthat

theiruseisbasedupontheerroneousassumptionthat‘thereisanaturalwayof

sayinganythinginanylanguage,andthatallweneedtodoistofindouthowto

saysomethingnaturallyinlanguageAandlanguageB’.

Thesubcorporainabilingual(ormultilingual)corpusmay,ontheotherhand,be

relatedthroughtranslation,thatis,thecorpusmaycontaintextsinone

language,alongsidetheirtranslationsintoanotherlanguage(orotherlanguages).

Suchcorporaarecommonlyknownasparallelcorpora,althoughtheterm

translationcorpusisalsoused(AltenbergandGranger2002).Parallel

corporaareusuallyaligned(Véronis2000).Thatis,explicitlinksareprovided

betweenunitsofthesourceandtargettexts,usuallyatthesentencelevel.This

enablesbilingualconcordancing,whereasearchforawordinonelanguage

returnsallsentencescontainingthatword,alongwiththeiralignedequivalent

sentencesintheotherlanguage.Parallelcorporaexistforseverallanguagepairs/

groupsoflanguages.Someareabyproductofbilingualormultilingual

parliaments:theEnglish–FrenchHansardsinCanada(ChurchandGale1991)

andthemultilingualEuroparlcorpus(Koehn2005),whichcontainsthe

proceedingsoftheEuropeanParliament,aretwowellknownexamples.Other,

morehandcrafted,parallelcorporaarecreatedspecificallyforuseintranslation

studiesandcontrastivelinguistics,andanumberofvariationsonthebasic

designarepossible:abilingualparallelcorpuscanbeunidirectionalorbi

directional,forinstance.GiventhatbidirectionalcorporasuchastheEnglish–

NorwegianParallelCorpus(Johansson1998)containsourcetexts(or

‘originals’)inbothlanguages,theycanalsobeusedasbilingualcomparable

corpora,provided,ofcourse,thatconditionsofcomparabilityobtain.Other

parallelcorporamaycontain,ontheirtargetsides,twoormoretranslationsinto

thesamelanguageofthesamesourcetext(Winters2005),orprogressive

draftsoftheemergingtargettext(Utka2004).Parallelcorporahavebeenused

intranslationTRAININGANDEDUCATIONtosupportstudentsinfinding

solutionstoproblemsthatcharacteristicallyariseintranslationbutnotother

sortsofwriting(Pearson2003),andinresearchintotranslationSHIFTS

(Munday1998a,2002).Theyhavealsobeenusedfortheextractionofde

factotranslationequivalentsinbilingualterminographyandlexicography

(BowkerandPearson2002:171–2;Teubert2002,2004),andtoprovide

empiricaldataforcorpusbasedMACHINETRANSLATIONsystems

(Hutchins2005a).

Corpusbasedtranslationstudies

MuchearlyworkinCTSsetouttopursuetheresearchagendaputforwardin

Baker’sseminal1993articleandinvestigated,onascalethathadnotbeen

possiblebefore,thoserecurrentfeaturesthatwerethoughttomaketranslation

differentfromothertypesoflanguageproduction.Thesefeatures,alsocalled

UNIVERSALSoftranslation,includedthereportedtendencyoftranslated

textstobemoreexplicit,usemoreconventionalgrammarandlexis,andbe

somehowsimplerthaneithertheirsourcetextsorothertextsinthetarget

language.Muchofthisworkwasconcernedwithoperationalizingabstract

notionslikesimplificationandEXPLICITATION(see,especially,Baker

1996a),andwithinvestigatingthepotentialofthequantitativetechniques