机器口译和人工口译衔接手段偏移对比研究_.pdf

ABSTRACT

Withthedevelopmentofinformationtechnology,artificialintelligencehas

becomeahottopicinrecentyears.AlthoughAIinterpretationapplicationshave

broughtgreatconveniencetopeople’scross-culturalcommunication,thereisstilla

gapinitsqualitycomparedwiththatofthehumaninterpreter.Aspreviousresearchers

focusonageneralassessmentofinterpretationqualityinfidelity,expressionand

grammar,thisstudyisconductedfromaspecificperspective:shiftsofcohesive

devices.

Inpreviousstudies,discoursecohesionisanimportantcriterionforinterpretation

qualityassessment.Asthecohesivetiesofacoherenttext,cohesivedevicesprovidea

clearwaytoevaluatetextcohesion.Bycomparingthedifferencebetweenhumanand

machineinterpretationsinshiftingcohesivedevices,theauthorcouldobservetheir

respectiveemploymentofcohesivedevicesandanalyzetheirinterpretationquality.

Inthisstudy,theauthorchoosesHuawei’s2022annualreportpressconference

astheresearchmaterial.UndertheguidanceofHallidayandHasan’sCohesion

Theory,theauthoraddsannotationstothefourtranscribedtextsofthespeaker,the

interpreter,andtwomachinetranslators:YoudaoandSougou.Aftercorpustagging,

dataiscollectedforquantitativeandqualitativeanalysis.UsingCorder’sError

Analysis,theauthorcomparesmachineandhumaninterpretationsandconcludesthe

reasonsfortheirgap.

Theresearchfindingsareasfollows.Firstly,therearethreetypesofshiftsin

humanandmachineinterpretations:retention,omission,andsubstitution.Secondly,in

termsoflexicalcohesion(excepthyponymsandantonyms),reference,omission,and

conjunction(excepttemporalandadversativeconjunctions),machinetranslators

retainmoreandsubstituteoromitfewercohesivedevicesthanthehumaninterpreter.

Thirdly,thereasonsforthesedifferencesarethatmachinetranslatorsrelyheavilyon

thetranscribedtextsofspeechrecognition,whichhelpsthemr