統(tǒng)計機(jī)器翻譯中語料選擇方法研究

1、黑龍江大學(xué)碩士學(xué)位論文統(tǒng)計機(jī)器翻譯中語料選擇方法研究姓名：李函章申請學(xué)位級別：碩士專業(yè)：計算機(jī)應(yīng)用指導(dǎo)教師：韓習(xí)武20100501黑龍江大學(xué)碩士學(xué)位論文AbstractAmongthetranslationmethodsthatarebasedonstatistics，thescaleandqualityofthebilingualcorpus，aswellastheaccuracyofwordalignmentalgorithmhav

2、esignificantinfluenceontherunionoftranslationsystemLargescalelinguisticcorpusCanenhancetheaccuracyofword—alignmentalgorithmandimprovethefunctionofthesystem，butatthesametimeinsacrificeofincreaseintheloadofthesystemTherefo

3、recurrentlytheresearchinthestatisticalmachinetranslationsshouldnotmerelyblindlyenlargethequantityoftheexistingcorpus，inordertofortifythefunctionofSMTsystemTosolvetheproblemsabove，thispaperbringsforwardthesolutionsasfollo

4、ws：AmelioratingthecapabilityofcomputerbasedtranslationbybeReremploymentofexistingparalleltraimngcorpusWepreprocessthecorpusviaadaptingthemachinestudy；distinguishthecorpustoliteraltranslationpairsandfreetranslationpairsTh

5、ispaperinnovativelyclassifiestrainingdatawiththemethodwhichcombinesgrammaticalcompatibilitywithlexicalcompatibilityandimprovethemethodthattoimprovecrosslingualwordkernelwithlinguisticsknowledge，andimproveSMTmodelbyclassi

6、fyingcorpustrainingExperimentsshowthatliteraltranslationmatchestheintellectuallevelofSMTsystemmoreperfectlyIftremendouswordsaregainedinalargeamountofcorpustranslation，weightedtrainingdatecouldboostthetranslationcontribut

7、ionDetailedanalysisshowsthat，ifsomeheuristicpruningalgorithmisusedtoavoidtheincensementoftheprobabilityofOOVliteraltranslmioncorpuswillexertmorecontributionontheSMTmodelKeywords：StatisticalMachineTranslation(SMT)；Bilingu