搜索
1210093210093MachineTranslation:Past,Present,futureDaiXinyu,YinCunyan,ChenJiajunandZhengGuoliang(StateKeyLaboratoryforNovelSoftwareTechnology,DepartmentofComputerScience&TechnologyNanjingUniversity,Nanjing210093)AbstractThispaperfirstlypresentsthehistoryofmachinetranslation,andintroducessomeclassicalparadigmsofmachinetranslation:RBMT,SBMTandEBMT.Secondly,weintroducetherecentresearchonmachinetranslation,anddescribethehybridstrategiesonmachinetranslationindetail,anddiscusstheapplicationsofmachinelearningformachinetranslation.Wealsoanalyzethecurrenttechniquesaboutevaluationonmachinetranslation.Finally,wedrawaconclusionandprospectontheresearchofmachinetranslation.KeywordsMachineTranslation,RBMT,SBMT,EBMT,HSBMT,MachineLearning.18632001AA114102,2002AA117010-0440BoothWeaver5060ALPAC50WeaverChomsky50[1][2]2.1(Rule-BasedMachineTranslation)ChomskyChomskyMETALSYSTRANATHENEHY1/[3][4]2.2(Statistical-BasedMachinetranslation)Isabelle19873.2IBMBrown[1]1990tstˆbest-translationT=argmaxTfluency(T)faithfulness(T,S)TSfluency(T)faithfulnessT,S[27]Brown[5]EM[9,10][26][6]LanguageModelP(t)TranslationModelP(s|t)Decoder)|(maxargˆstPtt=seesawIBM[7]FranzJosef[8]ACL’022.3(Example-BasedMachineTranslation)Nagao[2]TranslationMemoryOntologyFederica[11]2002Gale[9]Kay[10]Satoshi[12]Template-based[13]3.1HybridStrategiesMachineTranslationnFrederking[14]PanglossMarkIIIchartPanglossBrown[28]N-GramPanglossNomoto[25]Satoshi[12]3.2[15][16][17][18]K[19][4][20].[21]J.Tenni[22]Charniak[4]Brown[5]NaïveBayesNeuralNetworkKNNDecision-TreesRuleLearningbasedonerror-driventransformationHMMMaximumEntropy,BootstrappingCo-TrainingConditionalRandomFieldsIBMBLEU[23]NISTNIST[24]BLEUN-GramNNBLEUIBMBLEUNISTBLEUNNNNISTNISTALPAC[1]P.F.Brown,J.Cocke,S.A.Della,et.AStatisticalApproachtoMachineTranslation,ComputationalLinguistics,1990,16(2):79-85[2]Nagao.M.AFrameworkofamechanicaltranslationbetweenJapaneseandEnglishbyanalogyprinciple,ArtificialandHumanIntelligence,1984,173-180[3]M.Kay.Unificationgrammar,TechnicalReport,XeroxPaloAltoResearchCenter,1983[4]EugeneCharniak.StatisticalLanguageLearning,Cambridge,MA:MITPress,1993[5]P.F.Brown,S.A.Della,L.M.Robert,et.TheMathematicsofStatisticalMachineTranslation:ParametersEstimation,ComputationalLinguistics1993,19:263-311[6]NieBen,S.Vogel,H.Ney,et.ADPBasedSearchAlgorithmforStatisticalMachineTranslation,1998,ACL36/COLING17:960-967[7]Y.Wilks.CorporaandMachineTranslation.InProceedingsofMachineTranslationSummitIV.137-146,Kobe,Japan,1993.[8]FranzJosef,HermannNey.DiscriminativeTrainingAndMaximumEntropyModelsforStatisticalMachineTranslation,Inproceedingofthe40thACL,295-302,Philadelphia,2002[9]Gale,WilliamA.,K.W.Church.AProgramForAligningSentencesinBilingualCorpora.Inproceedingofthe29thACL,177-184,1991[10]Kay,Martin.TextTranslationAlignmentComputationalLinguistics1993,19,121-142[11]Federica,Riccardo,Paolo.SearchingSimilarSentencesforEBMT,SEBD’02,Italia,2002[12]Satoshi,Francis,Yamato.AHybridRuleandExample-basedmethodforMachineTranslation.NLPRS-97,1997[13]KevinMaTait.TranslationPatternExtractionandRecombinationforEBMT,PH.Dthesis,UMIST,2001[14]Frederking,Nirenburg.ThreeHeadsareBetterthanOne,InproceedingofANLP-94,Germany,1994[15]ThorstenBrant.TNT–AStatisticalPart-Of-SpeechTagger.Inproceedingsofthe6thAppliedNaturalLanguageProcessingConference,Seattle,Washington,USA,2000[16]EricBrill.ACorpus-basedApproachtoLanguageLearning.PH.DThesis,UPENN,1993[17]WalterDaelemans,JakubZavrel,PeterBerck,et.MBT:AMemory-BasedPartofSpeechTaggerGenerator.InProceedingsoftheWorkshoponVeryLargeCorpora,Copenhagen,Denmark,1993[18]AdwaitRatnaparkhi.AMaximumEntropyModelforPart-Of-SpeechTagging.InproceedingsofEMNLP-96,Philadelphia,PA.1996[19]Ng,HweeTou,HianBengLee.Integratingmultipleknowledgesourcestodisambiguatewordsense:AnExample-BasedApproach.IntheProceedingsofthe34thACL,40-47,1996[20]Magerman,D.M.StatisticalDecision-TreeModelsforParsing.IntheProceedingsofthe33rdACL.Cambridge,MA,26-30,1995[21]RatnaparkhiA.ALinearObservedTimeStatisticalParserBasedonMaximumEntropyModels,InEMNLP2,Providence,RI.,1997[22]J.Tenni,A.Lehtola,et.MachineLearningofLanguageTranslationRules.IEEEInternationalConferenceonSystems,Man,andCyberneticsTokyoInternationalForum,Tokyo,JAPAN,1999[23]Kishore,Salim,Todd,et.,Bleu:aMethodforAutomaticEvaluationofMachineTranslation.IBMResearchReport,2001[24]AutomaticEvaluationofMachineTranslationQualityUsingN-gramCo-OccurrenceStatistics.NISTReport,2001[25]TadashiNomoto,PredictiveModelsofPerformanceinMulit-EngineMachineTranslation,MTSummitIX,2003[26]Chen,StanleyF.,JoshuaGoodman.Anempiricalstudyofsmoothingtechniquesforlanguagemodeling.TechnicalReportTR-10-98,CenterforResearchinComputerTechnology,HarvardUniversity,1998[27]EugeneCharniak,KevinKnight,KenjiYamada,Syntax-basedLanguageModelforStatisticalMachineTranslation,MTSummitIX,2003[28]R.Brown,Frederking.ApplyingstatisticalEnglishLanguageModellingtosymbolicmachinetranslation.TMI’95,221-239,Leuven,Belgium,1995.