基于相似性英漢機(jī)譯系統(tǒng)的研究與實(shí)現(xiàn).pdf

1、東北大學(xué)碩士學(xué)位論文基于相似性英漢機(jī)譯系統(tǒng)的研究與實(shí)現(xiàn)姓名：張躍申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)軟件指導(dǎo)教師：姚天順2000.1.1壅j!、大堂堡主堡壅壘!!堅(jiān)!墮AbstractAtechnologyofMachineTranslationbasedonbilingualcorpusandsimilarlyreasoningisdiscussedinthispaperThismethodembodiestheassumptionthat

2、humanlanguageperceptionandproductiondependonconcretepastIanguageexperiences，ratherthanooabstractgrammarrulesThenamodelsystemofSimilarityBasedMachineTransiation(sBMT)ismadebasedonthecombinationofsimilaritytechnologywithEx

3、ample—BasedMachineTranslation(EBMTlafterthediscussionofthefeasibilityoftheapplicationofsimilarityintheMachineTranslationIntheSBMTmodelthenewterms“PARAWORD”“PARASENTENCE”and“PARASENTENCETRANSLATIONMOULD(PTM)”inthesentence

4、ofsourcelanguagearedefinedfortheexpressionconvenienceWealsodiscusshowwecanidentifyparawordinthesourcelanguagesentence，whetherthesourcelanguagesentencesareparasentences，whatisparasentencetranslationmouldorPTMhowtodrawoutt

5、hePTMfromthecertainsourcelanguagesentenceandhowtoaligntheanalyzedPTMw|Ihthepq“Msinthelibrary1nordertojncreasethePTMmatchingratiothepartialanalyseisusedtosubstitutethewholeanalyseandanewconce：ption，chunk，isreferredThealgo

6、rithmonidentifying，matchingandcombiningchunks，whichcanbeexpressedliterallyorinsimplegrammarrulesarediscussedindetailedinthispaperIntheconvenlionalalgorithms，themoretheanalyzingstepsofthesourcelanguagesentencestheloweriIs

7、analyseveracitywillbegenerallyInordertodrawoutthePTMoftbesoureelanguagesentence，nottheparsingtree，SBMTsystemonlyanalyzesitwhennecessaryThenthePTMwillbeusedtomatchwiththeonesinthePTMlibraryIfSUCCESSful，thetranslationofthe

8、source1anguagesentencewilfbegenerated，Otherwise，aPTMwithdeeperIevelwillbedrawnouttomakeanothermathingtry0fcoursethedeeperthePTMis，thelowerthereliabilityofthetranslationThefollowingcharacteristicswillbeshowninthefinalSBMT

9、system：1acertaintyfactorwillproducedwiththetranslationofsourcelanguagesentencewhichshowstheextenloilwhichusercarldepend；2repeatedtranslationwillbeavoidedcompletelywithhelpofthetranslationmemory；3SBMTcarlIearnthenewPTMfro

10、mitsuserswhichcanbeusedtomatchandreasonInSBMTsystemthecomputerandtheuserisawholeunitwhichcanhelpeachotherThestorageunitsofcomputers，whosecapacityisverylarge，canhelptheuserstememberthedataexactlyTheknowledgeinthelibraryca

11、nbereinforeedthroughtheuser’sstudyingThemostimportantfactoristhatthecomputercanprovidetheuseralotofpossiblyrelativedata，whichcaninspiretheuser’screativitybesidesthesimplereasoningConsideringthegeneralizationandpracticabi

12、lity，EnglishisselectedassourcelanaugewhileChineseastargetlanguageNowthemodelsystemcanworkwellunderthehelpofthesmall—scalebilingualcor!ousandcorrespondingPTMlibrariesTheEnglishsentenceinput|edcanbetranslationwithahighcert

13、aintyfactorifonlythereexiststheexampleinthebilingualcorpusorIhePTMwhichcanbematchedwiththeonedrawnoutfromtheinputtedsentenceMeantime，evenifthetranslationfailsthesimilarexamplesinthebilinguaIcorpuswillbeselectedandprovide