朴素贝叶斯论文：基于朴素贝叶斯模型研究语言特征对情态动词MUST语义影响【中文摘要】语义排歧研究是自然语言处理领域的热门话题,而情态动词研究是语言学领域里具有相当历史性而且很重要的研究方向。因此,利用自然语言处理的方法研究情态动词的语义是一个跨学科的研究,而且具有重要的意义。本文采用语义排歧方法来探讨情态动词must的语义问题。本文采用语料库定性定量相结合的研究方法和朴素贝叶斯有导语义排歧方法,以Coates(1983)对情态动词的研究成果为基础进一步探讨了如下两个问题:第一,不同语言特征对情态动词must的语义影响如何;第二,哪些语言特征相组合能产生最佳排歧效果。本文的创新点可体现在如下几个方面:第一,实现了朴素贝叶斯方法对情态动词must的语义排歧,使得语义排歧研究突破了目前大多只排歧名词、动词、形容词等语义比较简单的词的局限,从而使语义排歧研究迈进到了情态动词层面;第二,通过间接性的使用互信息(MI)值,使得互信息(MI)值这个重要的语言特征能在朴素贝叶斯排歧方法中发挥作用;第三,语言特征选择过程中条件互信息的使用,使得本文结论更具有科学性。实验结果显示,不同语言特征对情态动词must的语义影响程度也不同,按影响程度由大到小的顺序排列为:情态动词后动词的完成体;...【英文摘要】WordSenseDisambiguation(WSD)isahottopicofthefieldofNaturalLanguageProcessing(NLP),whilethestudyofmodality,whichhasalonghistory,isaveryimportantbranchoflinguisticfield.Therefore,thesemanticstudyofmodalverbsbythemethodsofNLPisacross-disciplinarystudyandofgreatsignificance.ThisthesisadoptstheWSDmethodtoconductasemanticinvestigationintothemodalverbmust.Thisthesistakesacorpus-basedqualitativeandquantitativeresearchmethod.Basedon...【关键词】朴素贝叶斯语义排歧情态动词must语言特征【英文关键词】Na(i|¨)veBayesWordSenseDisambiguationmodalverbmustlinguisticfeatures【索购全文】联系Q1：138113721Q2：139938848【目录】基于朴素贝叶斯模型研究语言特征对情态动词MUST语义影响摘要5-7Abstract7-8Chapter1Introduction15-201.1Backgroundofthisstudy15-171.2Significanceandpotentialapplicationofthepresentstudy17-181.3Objectivesofthepresentstudy181.4Creativepointsofthepresentstudy181.5Outlineofthepresentstudy18-20Chapter2LiteratureReview20-332.1Studyofmodality20-232.2StudyofWSD23-272.3StudyofNBinWSD27-302.3.1ApplicationofNBinWSD27-292.3.2ExplanationofthesuperbperformanceofNB29-302.4Briefdiscussionaboutthepreviousstudies30-322.5Summary32-33Chapter3ResearchMethodandDataCollection33-373.1Researchmethod33-353.2Datacollection35-37Chapter4WSDbyNa?veBayesianModel37-654.1Theoreticalfoundations37-504.1.1FundamentalknowledgeaboutWSD37-424.1.2Theoriesonmodality42-454.1.3Coates’viewsaboutmodalverbMust45-474.1.4Na?veBayes47-504.2DesignofNBWSDmodel50-544.2.1Determinationofthesenseinventory504.2.2DeterminationofthetraininginformationsourceforWSDmode50-514.2.3Featureextractionfromthecontext51-524.2.4DeterminationoftheWSDmethod52-534.2.5DeterminationoftheWSDalgorithm53-544.2.6Evaluation544.3ExperimentsbyNBWSDmodel54-644.3.1TrainingoftheNBWSDmodel54-604.3.2TestingoftheNBWSDmodel60-644.4Summary64-65Chapter5ResultAnalysisandDiscussion65-865.1Resultanalysis65-745.1.1Resultanalysisofthesolutiontothefirstproblem65-725.1.2Resultanalysisofthesolutiontothesecondproblem72-745.2Discussion74-855.2.1Discussionsabouttheconclusionsof5.174-805.2.2Discussionsaboutthecorpusesinthisstudy80-815.2.3DiscussionsaboutthesoundnessoftheindirectuseofMIinthisstudy81-825.2.4Discussionsaboutthefeatureofexistentialsubject825.2.5Discussionsabouttheexperimentsofrankingfeaturesandfindingthebestfeaturecombination82-835.2.6Comparisonwiththeresultsofpreviousrelatedstudies83-855.3Summary85-86Chapter6Conclusion86-906.1Mainfindingsoftheresearch86-876.2Implicationsofthepresentstudy87-886.3Limitationsofthepresentstudy88-896.4Furtherresearchsuggestions89-90References90-98AppendixⅠ98-109Acknowledgements109-110作者简介110
