《基于依存树库的英语名词句法研究.pdf》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。
⑧论文作者签名：指导教师签名：论文评阅人1：评阅人2：评阅人3：评阅人4：评阅人5：答辩委员会主席：一隧疆垄熬拯∑逝江王直太堂一委员1：隆盟受熬援∑逝洹工直太堂委员2：奎徨直到熬拯∑逝江太堂委员3：壁送壅熬拯∑逝江太堂委员4：委员5： ExaminingCommitteeChairperson：Dateoforaldefenee：27／05／2011凼：耋I一幽业一L!||}艺{|l～一～一一～～一一一 浙江大学研究生学位论文独创性声明本人声明所呈交的学位论文是本人在导师指导下进行的研究工作及取得的研究成果。除了文中特别加以标注和致谢的地方外，论文中不包含其他人已经发表或撰写过的研究成果，也不包含为获得逝壅太堂或其他教育机构的学位或证书而使用过的材料。与我一同工作的同志对本研究所做的任何贡献均己在论文中作了明确的说明并表示谢意。学位论文作者繇铂咱签字眺圳年6月7日学位论文版权使用授权书本学位论文作者完全了解浙江太堂有权保留并向国家有关部门或机构送交本论文的复印件和磁盘，允许论文被查阅和借阅。本人授权逝江太堂可以将学位论文的全部或部分内容编入有关数据库进行检索和传播，可以采用影印、缩印或扫描等复制手段保存、汇编学位论文。(保密的学位论文在解密后适用本授权书)靴敝雠轹唯必即新鹤：袈瑶签字日期：弘1)年占月7日签字日期：弦11年6月8日 摘要如今，对生语料进行句法标注己成为语料库语言学的主流趋势。树库，即通过词性和句法标注过的语料库，作为获取句法结构的知识源和评估句法分析结果的双重工具，引起了理论语言学和计算语言学学者们的浓厚兴趣和广泛重视。树库所含的大量词性句法功能分布信息亦为词性语法功能的理论研究提供牢靠的事实论据。本研究在概率配价模式理论基础上，利用英语依存树库，量化分析英语名词的各依存关系，通过逐层解析四个研究问题来描述英语名词的搭配和句法功能，挖掘本研究和其他中英文研究成果的同异之处。本研究的语料取自新概念英语和牛津高中英语课本，整个依存树库由6200条英语句子依存语法标注构成，包含句中各个词的词性，支配词，从属词及相应依存关系等信息。该树库的语法标注工作结合了斯坦福分析器的自动语法分析和人工校对两个部分，全部数据都被自动输入进Excel表格，然后应用Excel分别对名词作为支配词或从属词两种情况进行名词句法功能的统计分析，最后根据名词各句法功能出现的频率高低来区分名词的主要、次要和罕见句法功能。研究结果反映，名词的主要句法功能大体上证实了前人理论，更在极大程度上强调了主、次要句法功能的分布和使用，并补充了一些罕见句法功能。最后在总结的同时，提出了对现行国际上通用的英语依存句法标注体系的一些补充建议。关键词：语料库语言学，句法标注，依存语法，树库，概率配价模式，斯坦福语法分析器，英语名词，量化分析 AbstractNowadays，thesyntacticannotationonrawcorporabecomesadominanttendencyincorpuslinguistics．AstheSOUrCeofobtainingsyntacticstructureandfoundationofestimatingthesyntacticparsing，treebanks，annotatedcorpus，havebeenattachedgreatImportancebythescholarsfromtheoreticalandcomputationallinguists．Thelargeamountofdistributionalinformationonsyntacticfunctionofpart-of-speechcanmakeenormouscontributiononthetheoreticallinguisticresearch．011thebaseofPvP(ProbabilisticValencyPattern)theory,thisresearchaimstoclarifythecommoncollocationsandsyntacticfunctionofnouns，makeatomparisonwithotherpreviousresearchesviaaquantitativeanalysisofEnglishdependencytreebankandrespectiveinterpretationsoffourresearchqueStions．Therawmaterialsofover6200EnglishNewConceptEnglishandAdvancewithsentencesaleextractedfromthetextsofEnglish，andthedependencytreebankcontainspart。of-speechtagsandsyntacticstructuretagsofthesentences，includingtagsofthegovemor’dependentofeachword，andthecorrespondingdependencyrelations．Theannotationoftreebankwascompletedintwophases：theautomaticparsingofStanfordparserandmanualverificationofthepreviousresults．AllthesedatawereautomaticallygeneratedintotheExcelform，andthequantitativereportofthesyntacticfunctionswasanalyzedbyExcelinthelightoftwoconditionswhereEnglishnouDsactthegovemoranddependentdistinctively．Aftertheseriesofworkdonepreviously,thetypicalandatypicalsyntacticfunctionsofEnglishnounsWOuIdbeconcludedaccordingtothefrequency． Keywords：corpuslinguistics，syntacticannotation，dependencygrammar,Weebank,PVPtheory,Stanfordparser,Englishnouns，quantitativeanalysis ThesisforMasterDegreeofZhejiangUniversityListofT曲lesContents摘要⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．iAbstract⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯iiChapterOneIntroduction⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯11．1BackgroundofResearch⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯I1．2SignificanceofResearch⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．31．3PurposeofResearch⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯51．4OrganizationoftheThesis⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯5ChapterTwoLiteratureReview⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯72．1DependencyGrammar⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯72．1．1PhylogenyofDependencyGrammar⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．72．1．2VarietiesofDependencyGrammar⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．112．1．2．1WordGrammar⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯122．I．2．2FunctionalGenerativeDescription⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯⋯．．122．1．2．3DependencyUnificationGrammar⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯122．1．2．4Meaning—TextTheory⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．132．1．2．5ConstraintDependencyGrammar⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．142．2CorpusLinguisticsandTreebank⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．142．2．1CorpusLinguistics⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．152．2．2Treebank⋯⋯⋯⋯．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯⋯⋯⋯⋯．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．162．2．3ComparisonofDependencyStructureandPhraseStructure⋯⋯⋯⋯⋯⋯⋯⋯⋯172．3PVPtheory⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．182．3．1DefinitionofValency⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯182．3．2PVPTheoryanditsRelatedResearch⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．20ChapterThreeResearchMethodology⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．253．1ResearchQuegions⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯253．2Materials⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯．⋯⋯⋯⋯⋯⋯．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯253．3Instruments⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯26IV ThesisforMasterDegreeofZhejiangUniversityListofTables：；．3．1Parser⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯273．3．1．1TypesofParser⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯。⋯⋯⋯⋯⋯⋯．．283．3．1．2StanfordDependencyParser⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯29：；．3．2Excel⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．3l3．4ResearchProcedures⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．323．4．1ConstructingtheCorpus⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．323．4．2AutomaticPart-of-SpeechTagging⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯333．4．3AutomaticDependencyRelationsTagging⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．333．4．4ManualDependencyRelationsChecking⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯⋯⋯．．-⋯333．4．5StatigicalAnalysis⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．34ChapterFourResultsandDiscussion⋯．⋯．⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯⋯⋯．⋯．⋯⋯⋯⋯⋯．⋯．⋯354．1DependencyRelationsofEnglishNounsastheGovernor⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯354．1．1ProbabilityofDependencyRelationsGovernedbyEnglishNouns⋯⋯⋯⋯⋯．354．1．2GrammaticalFunctionsofEnglishNounsastheGovernor⋯⋯⋯⋯⋯⋯⋯⋯⋯．384．2DependencyRelationsofEnglishNounsastheDependent⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．414．2．1ProbabilityofDependencyRelationsGoverningEnglishNouns⋯⋯⋯⋯⋯⋯．414．3TypicalSyntacticFunctionsofEnglishNouns⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯444．4ComparisonbetweenEnglishNounsandChineseNouns⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．49ChapterFiveConclusion⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．531；．1Conclusions⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．535．2LimitationsandSuggestions⋯．⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯．⋯⋯．⋯⋯．⋯⋯．⋯⋯⋯⋯⋯⋯⋯．54References⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯56Appendix⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯⋯．⋯⋯⋯．⋯⋯⋯⋯⋯⋯⋯．⋯⋯．⋯．⋯⋯⋯⋯⋯⋯⋯⋯．．63AppendixI⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．63AppendixII⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯66AppendixIII⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．67Publications⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．68Acknowledgements⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．69V ThesisforMasterDegreeofZhejiangUniversityListofTabiesListofFiguresFigure2．1GraphicalRepresentationoftheSentence：Tomeatsanapple⋯⋯⋯⋯．8Figure2．2GraphicalRepresenantionoftheSentence：Myyoungbrotherlikesthisbeautifulgirl．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．8Figure2．3GraphicalRepresentationoftheSentence：Thisbeautifulgirlfasinatesmyyoungbrother．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．9Figure2．4DependencyRelationshipofPartofSpeech⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．9Figure2．5PhraseStructureTreefortheSentence：‘‘Economicnewshadlittleeffectonfinancialmarkets’’⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．18Figure2．6DependencyStructureTreefortheSameSentence．⋯⋯⋯⋯⋯⋯．⋯．⋯．18Figure2．7DiagramofPVPTheory⋯⋯⋯⋯⋯．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．⋯．⋯⋯⋯⋯⋯⋯．21Figure2．8DependencyRelationsRepresentationsofChineseVerbandNouninPVP⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯23Figure2．9FindingofChineseNounsastheGovernorinGaoSong’SStudy⋯⋯24Figure：；．1ScreenShotoftheTranscript．⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．26Figure3．2Auto—GeneratingExcelFileoftheSentence：Mrs．Smith’Skitchenissmall⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．26Figure3．3ProcessofParsingaSourceString(ComputerLanguage)⋯⋯⋯⋯．⋯．28Figure3．4GrammaticalRelationHierarchy⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．30Figure3．5ResultsofStanfordDependencyParserinExcelFormat⋯⋯．⋯．⋯⋯．．31Figure3．6StatisticalResultofExcel⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．32Figure4．1PVPGraphofEnglishNounsasGovernor⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．46Figure4．2PVPGraphofEnglishNounsasDependent⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯47VI ThesisforMasterDegreeofZhejiangUniversityListofT：lblesListofTablest出le4．1DependencyRelationsDistributionofNounsasGovernor⋯⋯⋯⋯⋯⋯36Table4．2PrepositionModifiersDistributionforEnglishNouns⋯⋯⋯⋯⋯⋯⋯⋯．391’able4．3DependencyRelationsDistributionofNounsasDependent⋯⋯⋯⋯⋯．41Table4．4SimilaritiesandDifferencesofPreviousStudyandPresentStudyonEnglishNouns⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．z18Table4．5SimilaritiesandDifferencesofChineseandEnglishNouns觞Governor⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．49Table4．6SimilaritiesandDifferencesofChineseandEnglishNounsasDependent⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯．．51Ⅶ ChapterOneIntroductionAstheintegrationoftime·-honoredlinguisticresearchandcutting··edgecomputerscience，naturallanguageprocessing(NLP)hasarousedgreatenthusiasmamongthewesternandChinesescholarsofvariousscientificdisciplinessince1950s．Before1990s，mostNLPsystemswerebasedoncomplexsetsofhand—writtenrules，whichmetseveraldifficultiesinlarge—scalerealisticcorpora．ArevolutiontowardscorpuslinguisticshadcameforwardafterCOLING1990(InternationalConferenceforComputationalLinguistics)andTMI1992(TheoreticalandMethodologicalIssuesinMachineTranslation)．Sincethat，parsedcorpora,commonlycalled‘treebanks’，commencesitsleadingperformanceinempiricallinguistics硒well硒inmachinelearningmethodsinnaturallanguageprocessing．Particularlyinlinguisticdomain，theusageoftreebanksisattachedmuchmoreimportancetothesyntacticresearchinrecentyears(Abeilld，2003；HinrichsandKiibler,2005；Nivre，2005；Hudson，2007；Liu，2009)．Phrasestructuregrammaranddependencygrammar(alsoknownasvalencygrammar)arethetwoessentialtheoreticaltoolsinsyntacticallyannotatingtherawcorpora．Althoughphrasestructuregrammar,theoutcomeofdominantChomsky锄theories，hasbeenwidelyappliedintoNLPdomainwithhigh—raterecognition，theincreasinginterestinusingdependency—basedrepresentationseemstobedrivenbothby‘'thepotentialusefulnessofbilexicalrelationsindisambiguationandbythegainsinefficiencythatresultfromthemoreconstrainedparsingproblem’’fortheserepresentations(Nivre，2005)，derivingfromthese，dependencyparsersareexpectedlyemployedtofulfillNLPtasks，frommachinetranslationtoquestionanswering．Inthelightofthistrend，dependencytreebank-basedlinguisticresearchleavesahotvacancytofillin．1．1BackgroundofResearchComparedwithphrasestructuregrammar,theresearchesandexperimentsondependencygrammarwerecarriedoutlaterinEnglishsyntacticparsing．Dependency IhesislorMasterDegreeofZhejiangUniversityChapmrOneIntroductiongrammar(grammairededependance)andstructuralsyntaxwasfirstlyformallyintroducedbyFrenchlinguistTesni6re(1959)．Mel’。cuk(1988)operatedthecomprehensiveinvestigation011dependencygrammarandhaddevelopedMeaning—TextTheory(MTD．Hudson(1984，1990)putforwardthenotionofWordGrammar．Hellwig(1986，2003)raisedthedefinitionofDependencyUnificationGrammar．Besidesthesewell-knowntheoriesofDependencyGrammar，agreatnumberofscholarsalsodevotedtheireffoastorefineandimprovethistheory(P．Sgall，1986；Pertsov，1987；Maruyama,1990；Karlsson，1990；Lafferty，1992；Milward，1994；HarperandHelzerman，1995；J"arvinenandTapanainen，1998；Krui3iif,2001；DuchierandDebusmann，2001；Schr"。oder，2002；CreswellandRambow，2003；Dras，2004；Klein，2004；Nivre，2005；Eisner,2005)．Dependencytreebankisatextcorpusinwhicheverysentencehasbeenannotatedwithdependencystructures．Eisner(1996)originallytransferredthePennTreebankintodependencyrepresentations，whichWaSnotedforphrasegrammarannotation。OntheresearchfoundationofMagerman，CollinsandYamada，McDonald(2005)andNivre(2006)perfectedthetaskoftransferringPennTreebankintodependencystructure．Withtheprevalenceofbuildingupadependencytreebank,dependencytreebankswithavarietyoflanguageversionshavebeenexploitedalltheseyears，forinstance，PragueArabicDependencyTreebank(PADT)andQuranicArabicDependencyTreebank(QADT)forArabic，atentativeChineseDependencyTreebank(exploredbyCommunicationUniversityofChina)forChinese，PragueDependencyTreebankforCzech，PragueEnglishDependencyTreebankandThePARC700DependencyBankforEnglish，ATRDependencycorpusforJapanese，LatinDependencyTreebankforLatin，ere．Duringthelasttwodecades，theapplicationofstatigicalmethodsinNLPhaSbeenproventriumphant．Manynaturallanguageprocessingsystemsaretypicalstatisticalmodelsderivedfromlargedatasetsapplyinglearningtechniques，e．g．，textnormalizers，part·of-speechtaggers，context-dependencymodels，acousticHidden-MarkovModels(Jurafsky&Martin，2009)．Especially,WeightedContext-FreeGrammar(WCFG)andProbabilisticContext-FreeGrammar(PCFG)aretwo2 !!!!堡!堕丝竺!竺里!￡苎竺21圣垒笪!竺墨望里!：!堡!垡竺皇!￡!竺虫!!!螋竺竺renownedgrammaticaltheoriesinphrasestructureparsing，betweenwhichthelatterisaspecializedformoftheformer．“Aprobabilisticcontext-freegrammarisacontext—grammarinwhicheachproductionisaugmentedwithaprobabilityoranumericweight，andtheprobabilityofaparseistheproductoftheprobabilitiesoftheproductionsusedinthatderivation”(Wikipedia)．Hence，thisprobabilitymodelcouldbeborrowedintothedependencygrammarstudy．Moreover,astatisticalsyntacticresearchonpartofspeechinsentenceorcomponentsofsentencewouldbeindispensiblefortheruleprogrammingofaparserorlexicalanalyzer；ontheotherhand，itwouldundoubtedlycomplementtheempiricallinguisticresearchandwordsgrammarfunction．FengZhiweiandLiuHaitao(2007)proposedaPvPtheory(ProbabilisticValencyPattern)，whichnotonlyextendedthetraditionalvalencygrammar,butalsoaddedthestatisticalmodelintovalencypatterninNLEBymeansofthistheory,itcouldbeobservedthatanypartofspeechcoulddominateorbedominatedbywhichkindofdependencyrelationsinqualitativemethods，andthefrequencyanddistributionofthesedependencyrelationsalsocouldbededucedbyquantitativemethods．WiththehelpofatentativeChineseDependencyTreebankandtheguidanceofPVPtheory,GaoSong(2010)hadaccomplishedtwocrucialquantitativestudiesonsyntacticfunctionsofChineseverbsandnouns．Thisstudygreatlyemphasizesandamelioratesthesyntacticfunctionsofthepartsofspeechwhichwereusedtobeinvestigatedbypurelytheoreticalmeans．IncontrasttoChinesedependencygrammar,countlessscholarswithdifferenteducationbackgroundhavefloodedintothestudyofEnglishdependencygrammar．However,almostallofthemwoulddemonstratetheirvigorandfocalspotontheapplicationofdependencystructuresinNLP,andtherearefewstudiessettledinthesyntacticdomain．Thus，theincorporationofconventionalframeworkandmoderntechniqueswouldbeinurgentneedinEnglishdependencyfield．1．2SignificanceofResearchDuringtheshortevaluationhistoryofNLP,amilestonewasfoundedin2006and2007withtheperformanceoflarge—scaleevaluationofdependencyparsersinthe3 。_一～～h一，一』‘_‘“H√一”。‘l’hesisforMasterDe帮eeofZheiiangUniversityChapterOneIntroductioncontextoftheC汛Lsharedtasks．ThismarkedthegeneralUSeofdependencygrammarinnaturallanguageprocessing,andalsoprevailingdevelopmentofdependencyparserinanalyzingtherawcorpora．Sincedependencygrammardefinesthateachwordclassdistributesunevenlyinthesentence，onespecificwordclasscoulddominateandbedominatedbyotherwordclassesthroughcertaindependencyrelations．Besidesthat，theoccurringfrequencyofeverydependencyrelationisquitedistinct,thusaquantitativeanalysisofdependencyrelationsshouldbeinvolvedintotheempiricallinguisticresearch，alsogrammaticalstudyofthewordclass．Atthesametime，theprocessingtaskoflarge-scalerealistictextshasbeenputmoreandmorcemphasison，andtheunanimousrealizationhasbeenreachedinthecomputationallinguisticsarea,corpusistheoptimalresourceoflinguisticknowledge．Moreover'annotatedcorpuswouldbethemostaccessibleonetobeputintoservice，andthisisexactlytheTreebank．AmongallthepartsofspeechinEnglishgrammar,nounsformthelargestwordclass．AccordingtoCarterandMcCarthy(2006)，nounsaredescribedastheclassesandcategoriesofthingsintheworld，includingpeople，animals，inanimatethings，places，events，qualitiesandstates．Nounscanalsobecreatedbyconversionofverbsoradjectives．InRandolphQuirk’S(1985)comprehensivegrammaroftheEnglishlanguage，thenounsandnounphrasestypicallyfunctionsassubject,object,complementofclausesandcomplementofprepositionalphrases，includingtherareconditionoffunctioningasadverbialmodifiers．Consideringthediversityandflexibilityofnouns’functioninthesentence，thestudy011nounswouldbetheelementarypaceofthewholegrammaticalworld．Mostofall，PVPtheoryisanew-bornbabygivenbirthbythetwooutstandingChineselinguistswhohadmadegreatachievementinmanydisciplines．RelatedresearcheshavebeenconductedinChinesegrammarlastyear．Quasi—Englishdependencygrammarresearchhasn’tbeencarriedoutyet．Builtonthestatisticalresultofdependencytreebank,thisresearchwouldverifythepreviousnormsofEnglishnounsbycomparingthedependencyrelationsoccurring4 ThesisforMasterDegreeofZhejiangUniversityChapterOneIntroductionfrequencyofnounsasthegovernorandsubordinate，forinstance，commonsyntacticalfunctions；alsoitwouldsearchoutsomeunusualevenundiscoveredfunctionofnounstosupplementthecurrentuniversalgrammar．ThetypicalandatypicalanalyzedsyntacticfunctionsofEnglishnounswouldhencemakefortheemphasisontheprimaryandcommonEnglishgrammarintheprocessofgrammarteaching．Itmeansthattheteacherwoulddiscriminatetheimportanceoffunctionsofcertainpartofspeechwhenhe／sheisgivingthelessonstothenativestudentsortheforeignstudentsstudyingforsecondlanguage．Apartfromthetheoreticalstudyonlinguistics，thesetypicalandatypicalfunctionsofwordswouldhelpimprovetheprogramminglanguageofaparser,andthemoFeaccuratedescriptionofaparserwouldenhancetheprecisionofparsingresults，whichenormouslyreducethecostofmanualwork．Inaddition，thistreebankcouldbeenrichedwithmoredependency-relationstagsorsemantic，phoneticinformationforfurtherresearches，andalsoitcouldbeanalyzedinotheranglestolearntheuseofotherpartsofspeechorotherlinguisticinformationofcertainpartofspeech．Allthesehaveproventhisresearchworthyofinvestigation．1．3PurposeofResearchBasedonthePvP(ProbabilisticValencyPattern)theory,thisresearchisaimedtoclarifythegrammaticalfunctionandsyntacticdistributionofEnglishnounsviaaquantitativeanalysisofEnglishdependencytreebank,whichcomprises6200Englishsentencestaggingofgovernor，subordinateanddependencyrelationsaccordingtothedependencygrammar．1．4OrganizationoftheThethesisiscomposedoffivechapters．Theintroductionpartconcernsthebackground，significanceandpurposeoftheresearch．FollowingChapterOneistheliteraturereviewpart．Thedetaileddefinitionandpreviousworkondependencygrammar,corpusandTreebanLPVPtheorywillbeinvolvedinChapterTwo．ChapterThreeisthemethodologypart，whichcentersontheresearchmethodandprocess，includingthedescriptionsofresearchquestions，materialsandinstruments．Foremost,theautomaticdatainputprogram，Stanforddependencyparseranddata与 堡箜!!塑竺竺堡!里!塑苎21圣!鱼!竺墨望竺堡!查垒=曼!!窆竺Q里!望业笪竺analyzingsoftwarewillbeallpresentedinthischapter．ChapterFourdepictstheresultsofdataanalysis，andtheanswersofthefourresearchquestionswillberevealedrespectively．Here，theproportiondistributionofdependencyrelationsofnounsasthegovernoranddependentwillbepresentedintables，withexplicitanalysisonsyntacticfunctionsofEnglishnouns．ThecomparisonbetweenChinesenounsandEnglishnounswillbeprovidedatthelastsectionofthischapter．TheconclusionreachedfromcurrentresearchisdrawninthelastchapteLthelimitationandprospectofthisresearchisalsolistedinChapterFive．6 ChapterTwoLiteratureReview2．1DependencyGrammarThefundamentalnotionofdependencyisbasedontheideathatthesyntacticstructureofthesentenceconsistsoftheasymmetricalsyntacticrelationsbetweenpairsofwordsofthesentence．ThisconceptiSofcentralimportanceintheoreticallinguistics，playingitsroleinalmosteverymajorsyntactictheory．2．1．1PhylogenyofDependencyGrammarTherootofdependencygrammarmaybetracedbacktoPanini’SformalgrammarofSanskritaround350to250BC．TheveryfirstsystemictreatmentofArabicgrammar,containingvariouslevelsoflinguisticinformation，wasconstructedontheconceptsthatformthecoreofnowadaysdependencygrammar．Inthissyntax,therewasanobviousdistinctionontheheadanditsdependents．InmedievaltimesofEurope，dependency—basednotionswereaudaciouslyusedbytraditionalgrammarians，particularlyThomasfromErfurtandMartinfromDacia．．Althoughsomepreliminaryspeculationshavebeendonebyancientandmedievallinguists，historicaldevelopmentsstartinginthe1950shavebroughtunprecedentedimpactontheformalgrammar．Andthesedevelopmentscouldbedividedintofourphases：formalization，addingmeaning，wordorderandgrammarconcerninglogic&computation张ruijif,2002)．WiththeprecedinginfluenceofBloomfield’Sdescriptivebehaviorismin1930sandGroundbreakingworkinmathematicallogic，computersduring1930sto1940s，NoamChomskypublishedhisfamousbookSyntacticStructuresin1957．Thisbookwaslaterwidelyconsideredasoneofthemostimportantworksinthefieldoflinguistics，andithadlaidthefoundationofhisideaoftransformationalgrammar．Healsodiscussedfinitestategrammarandphrasestructuregrammar,amodelfoundedonimmediateconstituentanalysis，whichhadmadegreatcontributiononNLParea．Ontheotherhand，anotherachievementonlinguisticsshouldbeattributedtoTesni色re，whohadbroughtthemodemnotionofdependencygrammarin1959．Inhis7 ．ThesisforMasterDegreeofZhejiangUniversity———————————————．．——．．．．．．．．．．．．．．．．C．h．a．p．．te．r．．T．w．．o．．L．i．te．r．a．t．u．r．e．．R．e．v．i．e．w．一●——．．．，．．．．．．．_-．．_．．．．．．．-●_-。_____。______________________________________________-。___。●___-。．．．●_-一．．．．．——masterpieceElementdeSyntaxeStructurale，heassumedthatsyntacticst加ctureisbasedonthethreecomplementaryconceptsofconnection(connexion)，junction(jonction)andtransfer(translation)，whereconnectioncorrespondstodependency,whilejunctionandtransferrefertootherkindsofrelationsbetweenthewordsofthesentence·Concerningthenotionofconnection，hebelievedthateverysentenceiscomposedbykindsofconnectionsbetweenthewordsofthesentence．Meantime．theconnectionbetweenwordsshouldobeythehierarchical(hi6rarchie)principle，inthisway,thedependencyrelationsbetweenwordscouldbebuiltupthroughtheseconnections·Agraphiccouldrepresentthisrelationship．Figure2．1givesthegraphrepresentationfortheexamplesentence“Tomeatsallapple”．Figure2．1GraphicalRepresentationoftheSentence：TomeatsallappleInthisgraph，theverb‘eats’istheknot(nQeud)ofthissentence，andthenouns‘Tom’and‘apple’aredependingontheverbwhichareplacedbelowtheverb：thearticle‘an’isdependingonthenoun‘apple’，beingplacedbelowthisword．fesni6reassumedthatsentenceisanorganicwhole，whosecomponentsarenotonlysignifyingthewordsonitssurface，butthesyntacticrelationsunderlyinginthesentence·Healsoconsideredverbsasthecoreofthesentenceandotherelements硒dependentsboundbytheverb’svalency．Inhistheory,thesubjectandobjectofthesentenceareregardedatthesamelevel，evencouldtheyreversetheirpositioninthisstructure．Figure2．2and2．3demonstrateacomparisonofthesetwosituations．／＼／＼young。／暇t如Figure2．2GraphicalRepresenantionoftheSentence：Myyoungbrotherlikesthisbeautifulgirl．8l驯p fasci燃tesf／／＼g缈brother一“／／＼。／＼b毫赡eautiful嘶两ungFigure2．3GraphicalRepresentationoftheSentence：Thisbeautifulgirlfasinatesmyyoungbrother．Adeductioncouldbederivedtothehierachicalprinciple，allthedependentswoulddependontheirgovernororregent，forthesubjectphraseofFigure2．2actsastheobjectofFigure2．3，whiletheobjectphraseofFigure2．2playsthesubjectofFigure2．3．Besides，junctionindicatestherelationbetweencoordinateditemsthataredependentSofthesameheadorheadsofthesamedependenLwhiletransferistherelationbetweenafunctionwordorotherelementthatchangesthesyntacticcategoryofalexicalelementSOthatitcanenterintodifferentdependencyrelations．Tesni6reproposedfourfundamentalpansofspeech：verb，noun，adjective，adverbial．TheirdependencyrelationscouldbeseeninFigure2．4asfollows：Verbp／No，一I汗陡A艇凼lal1砖她fe纠∥degree一3糟ck*gr蜊4砷aegr蜊Figure2．4DependencyRelationshipofPartofSpeechAsthegraphindicates，onlyverbisatthefirstdegree，nounandadverbialisplacedattheseconddegree，adjectiveandadverbialissetatthethirddegree，andatthefourthdegree，onlyadverbialcouldbefound．Afterwards，AmericanlinguistD．G．Hays(1960)deliveredallessayGroupingandDependencyTheory,andcameupwithameansofdependencyanalysis，whichoverlappedandimprovedTesni6re’Sdependencygrammartosomeextent．9 !堕!!121竺苎I竺旦!受121圣堕!竺曼竺翌!∑竺!壁里皇!￡坚!塑生!坚型堡壁：!竺Criteriaforestablishingdependencyrelationsandregulatingtherelationshipbetweentheheadanddependenthavebeendiscussedinthesubsequentyears．In1970，AmericancomputationallinguiStsJ．RobinsonissuedhisworkDependencyStructureandTransformationalRules，whereheraisedfouraxiomstoconstrainthedependencystructures：“1．Oneandonlyoneelementofthesentenceisindependent；2．Allotherelementsofthesentencedependoncertainelementdirectly；3．Noelementcoulddependdirectlyonmorethanoneotherelement；4．IfelementAdependsonelementBdirectly,whilesomeelementCintervenesbetweenelementAandB(inlinearorderofstring)，thenelementCcoulddependonelementAorB，orsomeotherinterveningelement．”Zwicky(1985)hadproposedsomerulesfordistinguishingasyntacticrelationbetweenaheadHandadependentD：“1．HdeterminesthesyntacticcategoryofCandcanoftenreplaceC：2．HdeterminesthesemanticcategoryofC；Dgivessemanticspecification；3．Hisobligatory；Dmaybeoptional；4．HselectsDanddetermineswhetherDisobligatoryoroptional；5．TheformofDdependsonH(agreementorgovernment)；6．ThelinearpositionofDisspecifiedwithreferencetoH．”K．Schubert(1987)submitted12principlesofdependencygrammardealingwithlanguageprocessingafterseveral—yearworkofengaginginDLT(DistributedLanguageTranslation)project．OnthefoundationofRobinson’Srules，heexpandedthetheoryintothefieldofmorphemeandcontext,andenhancedthecomputabilityandoperability．Thesetwotheorieswerewidelyadoptedinthenaturallanguageprocessingfield．InviewofChinesecharacteristicsandmachinetranslationpractices，FengZhiweiadvanced5conditionsforaD-tree，whichexplicitlydescribestherelationsbetweenthenodesandgreatlycomplementsthepreviousfindings．Amixofvariouscriteriaondependencyrelationsandthedifferenttypesofdependencyrelationsweresetbymanytheorists，suchasNikula,whobelieved‘thesyntacticdependencyshouldbedistinguishedinendocentricandexocentric10 !!!!!!堕坚竺坚里!塑竺21三堕!竺墨望呈!!!竖堕g!翌竺!塑兰!竺坐竺塑：!!∑constructions．Theheadcouldreplacethewholeelementswithoutdisruptingthesyntacticstructurein锄endocentricconstruction，anddependentsareselectedasoptional，notbytheirheads；whiletheheadcouldn’treplacethewholeelementseasilyinanexocentricconstruction．Thedistinctionbetweenendocentricandexocentricconstructionsisalsoregardedtobelinkedtothedistinctionbetweenhead—complementandhead-modifierrelationsincontemporarysyntactictheories，wheretheformerrelationsareexocentricwhilethelatterrelationsareendocentric．Thedistinctionbetweencomplementsandmodifiersisoftendefinedintermsofvalency,whichisacentralnotioninthetheoreticaltraditionofdependencygrammar’(Nikula,1986)．Inthefollowingsections，thedefinitionofvalencycouldbeexplainedmorespecifically．AccordingtoMel”cuk(1988)，thewordformsofasentencecouldbeconnectedbythreekindsofdependencies：morphological，syntacticandsemantic．Hudson(1990)suggested‘theconceptofgovernor／headhasprototypestructure，whichmeansthetypicalexamplesofthispartofspeechwouldmeetallormostofthecriteria,andtheatypicalonesmeetfewer’．Inconclusion，thetheoreticalnotionofdependencygrammarisintegratedwiththeassumptionthattheessenceofthesyntacticstructureofsentencesliesinbinaryasymmetricalrelationsbetweenlexicalelements．What’Smore，almostallimportantdependencytheoriesagreethatthereisacore／headinthesyntacticstructures．However,thereisalsosomecrucialdistinctionwithrespecttothetypesofsyntacticstructures，thecriteriasetfordrawingthedependencytrees，theformalpropertiesofdependencystructures，andSOon．2·1·2VarietiesofDependencyGrammarAlthoughthecentralideacouldbecommonlyreachedamongthesedependencyscholars，thereisstillsomedivergencefoundintheirderivationaltheories．Themostprevailingtheoreticalframeworksinclude：WordGrammar,FunctionalGenerativeDescription，DependencyUnificationGrammar,Meaning—TextTheory,ConstraintDependencyGrammar,WeightedFunctionalConstraintDependencyGrammar,FunctionalDependencyGrammar,Topological／ExtensibleDependencyGrammar,DependencyGrammarLogic，etc．Thedefinitionoffivecrucialvarietiesofll 堡!!!!堑坚兰竺竺里!塑竺21圣型i竺竖望里堡竺!堡兰!苎兰竺!迪生!望翌塾坚塑：!型dependencygrammarwillbedeliberatedasbelow．Z．1．Z．1WordGrammarTheconceptofWordGrammar(WG)wasputforwardbyRichardHudsonin1984．Itstartedasamodelofsyntax,whosemostdistinctivecharacteristicisitsuseofdependencygrammar,allapproachtosyntaxinwhichthesentence’Sstructureisalmostentirelycontainedintheinformationaboutindividualwords．andsyntaxisseen嬲consistingprimarilyofprinciplesforcombiningwords．Thecentralsyntacticrelationisthatofdependencybetweenwords；constituentstructureisnotrecognizedexceptinthespecialcaseofcoordinatestructures．HoweverallevenmoreimportantclaimofWordGrammaristhatstatementsaboutwordsandtheirpropertiesformacomplexnetworkofpropositions．MorerecentworkonWordGrammarcitesneuro-cognitivelinguistics嬲asourceofinspirationfortheideathatlanguageisnothingbutanetwork．OneoftheattractionsofthenetworkviewisthepossibilityofanalyzinglanguageinthesamewayaSotherkindsofknowledge，giventhatknowledge，orlong-termmemory,iswidelyconsideredtobeanetwork．WordGrammarisallexampleofcognitivelinguistics，whichmodelslanguage雏partofgeneralknowledgeandnotasaspecializedmentalfaculty．2．1．2．2FunctionalGenerativeDescriptionFunctionalGenerativeDescription(FGD)isalinguisticframeworkdevelopedatCharlesUniversityinPraguesincethe1960sbyateamledbySgall．BasedOilthedependencygrammarformalism，itis“astratificationalgrammarformalismthattreatsthesentence舔asystemofinter-linkedlayers：phonological，morphematical，morphonological，analytical(surfacesyntax)andtectogrammatical(deepsyntax)．ContinuingthetraditionofPragueSchool，specialattentionispaidtothephenomenonoftopic—focusarticulmion．ThePragueDependencyTreebank(PDT)isconsistingofasubsetoftheCzechNationalCorpusannotatedalongthelinesofFGD”．(Wikipedia)2．1．2．3DependencyUnificationGrammarIn1986，GermanylinguistHellwigfirstlyraisedtheconceptofDependencyUnificationGrammar,whichsharedtheconvergentprincipleswith12 !!!!!!塑坚呈壁竺里!塑竺21三竖!竺墨坚坐竺塑竺!璺苎竺堡2坠竺罂坚垦望!竺Government—BindingTheory,GeneralizedPhraseStructureGrammar,FunctionalUnificationGrammar：applythedependencyrelationsbyusingnotions‘Head／Governor’and‘Dependent’；attempttodemonstrationtherepresentationofsyntacticstructuresinfunctionalaspect；agreethatsyntaxshouldbelexicallyrestrictedandtogeneratelexicalinformationtothedeepstructure；setthealgorithmsontheprincipleofunificationandbringcomplexcategoriesintoagreementinthecontext．Hebelievedthat‘DUGscouldbeformulatedinthecomputerlanguageDRL(DependencyRepresentationLanguage)．Accordingtounificationgrammars，eachfeatureiSannotatedasanattribute．valuepair．DuetotheDUGpresumesthatadifferenceshouldbemadeamongthreedimensions：lexicalmeaning，syntagmaticfunctionandoutwardform，SOthreetypesofattributesareSOrtedineachDRL-term：alexeme，asyntagmaticroleandamorpho-syntacticdescription．Unificationprincipleisemphasizedinhistheory,whichhasbeenputintopracticeinDRL．Thefundamentalprincipleofthemechanismisthenotionofsubsumption．Heproposedtouseaunification—basedparser．TocorrespondtotheDUG，thePLAINparserisintroducedinhiswork—aprogramsystemfordependencyanalysisandforsimulatingnaturallanguageinference．2．1．2．4Meaning-TextTheoryMeaning-TextTheoryisthetheoryofthelanguagemadebyIgorMel’6ukwhointerpreteditasmulti—layermodeloftransformingtextintothemeaningandviceVerSa．TheMeaning-TextTheory(MTT)isatheoreticallinguisticframeworkfortheconstructionofmodelsofnaturallanguages．Startingfromthe1960s，theMTTsetitsfoundationsinMoscow(I．Mel’6uk,A．2；olkovskij，J．Apresjan)．SincethenithasbeengreatlydevelopedinRussia,andspreadtoCanadaandEurope．“TheMTTprovidesalargeandelaboratebasisforlinguisticdescriptionand，duetoitsformalcharacter,lendsitselfparticularlywelltocomputerapplications．Startingwiththetheoreticalprinciplethatlanguagesmustbedefinedbythewaytheirelementsarecombined．theMTTconsidersthatthelexicalunitsimposethiscombinationand，therefore，itiSinthedescriptionofthelexicalunitsandofits13 Helzerman，1995)CDGparseinformationcanbeeasilylexicalizedatthewordlevel．Thislexicalizationisabletoincludenotonlylexicalcategoryandsyntacticconstraints，butalsoarichsetoflexicalfeaturestomodelsubcategorizationandwh-movementwithoutacombinatorialexplosionoftheparametricspace(Wang&Harper,2002)．CDGcandistinguishbetweenadjunctsandcomplementsduetotheuseofneedroles(Harperand995)，ismorepowerfulthanCFG(Context-FreeGrammar)，andhastheabilitytomodellanguageswithcrossingdependenciesandfreewordorderingwhichcouldbeapplicabletoawidevarietyoflanguages．CDGsystemscanbeusedtocreatefullsyntactictreesinotherformalismsbyaddingsmall，non—terminalbaseddependencygrammars，andanumberofcorpus／treebankprojectshaveusedConstraintGrammarandConstraintDependencyGrammarforautomaticannotation．CGandCDGmethodologyhasalsousedinanumberoflanguagetechnologyapplications，such硒spellcheckersandmachinetranslationsystems．2．2CorpusLinguisticsandTreebankCorpusLinguisticshasbredahugenumberofresearchmethods，holdingtheirtenettotraceapathfromdatatotheory．ThecorpusapproachrunscountertoNoam14 (SEU)whichinitiatedthedevelopmentofcorpuslinguistics．AlandmarkinmoderncorpuslinguisticswasthepublicationbyHenryKuceraandNelsonFrancisofComputationalAnalysisofPresent-DayAmericanEnglishin1967，aworkbasedontheanalysisoftheBrownCorpus，whichwasacarefullycompiledselectionofcurrentAmericanEnglish，totalingaboutamillionwordsdrawnfromawidevarietyofsources．TheinitialBrownCorpushadonlythewordsthemselves，plusalocationidentifierforeach．Overthefollowingseveral—yearpart-of-speechtagswereapplied．ThetaggedBrownCorpusformedthebasisformanylatercorporasuchastheLancaster-Oslo／BergenCorpus．SuggestedbyProfessorGeoffreyLeechofLancasterUniversity,takenchargebytheProfessorStigJohanssonofUniversityofOslo，builtintheNorwegianComputingCentrefortheHumanitiesinUniversityofBergen，theLancaster-Oslo—BergenCorpus(oftenabbreviatedasLOBCorpus)in1980sprovidedaBritishcounterparttotheBrownCorpuscompiledbyKuceraandFrancisforAmericanEnglishin1960s．ItscompositionwasdesignedtomatchtheoriginalBrowncorpusintermsofitssizeandgenresascloselyaspossibleusingdocumentspublishedintheUKbyBritishauthors．Agroupoflarge—scalecorpusesdesignedforcompilingdictionaryemergedinthelate1980s．SuchasCOBUILD，anacronymforCollinsBirminghamUniversityInternationalLanguageDatabase，isaBritishresearchfacilitysetupattheUniversityofBirminghamin1980andfundedbyCollinspublishers．ThefacilityWaSledbyProfessorJohnSinclair．ThemostimportantachievementoftheCOBUILDprojecthasbeenthecreationandanalysisofacorpusofcontemporarytext(theBankofEnglish)，andtheproductionofthemonolinguallearner'sdictionaryCollinsCOBUILDEnglishLanguageDictionary(1987，2ndeditionin1995，3rdeditionin2001，4theditionin15 2003，5theditionin2006)，basedonthestudyoftheCOBUILDcorpus．Since1979，Chinahadmadegreateffortsonbuildingthemachine—readablecorpus．In1991，theStateLanguageCommissionhadstartedtobuildthestate—levelChinesecorpustopromotetheresearchesinChineseGrammar,Syntax,SemanticsandPragmatics，whichwasplannedtobethebiggestChinesecorpusincluding70millionChinesewords．Andtillnow,ithasfinishedtheinputandproofreadingof20millionwords．While，afewChineseinformationprocessingfirmsandorganizationshadindulgedintoconstructingthelargeauthentic-textcorpus：People’SDailyCD—ROMDatabase，InstituteofComputationalLinguisticsinPekingUniversity,TsinghuaUniversity,CityUniversityofHongKong，InstituteofSoftwareChineseAcademyofSciences，DependencyTreebankofChina’SCommunicationUniversity,SinicaofTaiwan，etc．ChinahadconstructeditsfirstEnglishcorpusinthemiddleof1980s：JDEST,whichwasmanagedbyYangHuizhongofShanghaiJiaotongUniversity．ThisEnglishcorpushadmadegreatcontributionontheestablishmentofCollegeEnglishCurriculumRequirementandvocabularystatistics．Later,otherEnglishcorpuseswercbuiltinChinasuccessively：ICLEChinaSub-corpus(GuiShichun)，CLEC(GuiShichun&YangHuizhong)，COLSEC(YangHuizhong)，SWECCL(WenQiufang)，CEC(LiWenzhong)，DMMEE，etc．2．2．2Treebankisadependencygrammartaggingcorpuswhichcombinesthewords’propertiestaggingandsentences’syntacticrelationstagging．Linguistscanmanuallyannotateeachsentencewithsyntacticstructuretocreatea仃eebank,ortheycanarrangeaparsertoannotatethesestructuresautomatically,andthentocheckoutthisworklatertocompletethistreebank．TheworldrenownedChineseTreeBanksincludeChinesePennTreebank(CTB)ofUniversityofPennsylvania,SinicaofTaiwan，andtheChineseTreebankofTsinghuaUniversity．Thetheoreticalandpracticalinvestigationofdependencygrammarreceivesincreasingrecognitioninrecentyears．EisnerfirstlytransformedthePennTreebankintodependencydescriptionforms，andthendidadependencysyntacticanalysis．Later,16 intodependencystructure；NivreandMcDonaldalsomadeaseriesofimprovementbasedonthepreviousresearches，whichtheyexploredatransformationalsoftwarePenn2Malt,andenhancedtheaccuracyofdependencyanalysis．Similarly,anumberoflarge·scaleEnglishtreebankswereexplored：Lancaster-LeedsTreebankandPennTreebank．Otherlanguagewell—knowtreebanksincludeNEGRAandTIGERtreebankofGerman，andthePragueDependencyTreebank(PDT)ofCzech．TreebankinmorelanguageversionscouldbefoundinChapterOne．2．2．3ComparisonofDependencyStructureandPhraseStructureUndeniably,mostoftreebanksprefertobelesstheory—specific，howeversometreebanksstillfollowaspecifictheoryintheirsyntacticannotation．Amongtheseloyalfollowers，twomaingroupsCallberecognized：treebanksannotatedwithphrasestructureandtreebanksannotatedwithdependencystructure．Thesyntacticrepresentationandfileformatwouldbeconstructedaccordingtoaparticulargrammar．AsFigure2．5and2．6showbelow,aphrasestructuretreeexplicitlyrepresentsthephrasesandstructuralcategoriesindicatinggrammaticalfunctions，andterminalnodesandlabelscouldn’tbefoundinthisstructure，andmakeexplicitsyntacticconstituents，suchasnounphrases(NP)，verbphrases(VP)andprepositionalphrases(PP)identifiableinthesentence；whiledependencystructuretreeclearlydemonstratesthehead·dependentrelations，functionalcategoriesbyapplyingthearclabels，andotherstructuralcategoriessuch嬲part-of-speech，andthearCSarelabeledindicatingthenatureofdependency．Inbothcases，eachsentenceisannotatedwithpart-of-speechinformation．Althoughdistinctionliesinthesetwostructures，hybridrepresentationandderivationalrepresentationofboththesetworepresentationsbecomemorefeasibletoadaptvarioussituations．Inthisresearch，onlydependencystructurewillbeemployedanddependencygrammarwillbeconsidered．17 nmods锄t≯Figure2．6DependencyStructureTreefortheSameSentence2．3PVPtheoryBeforeexploringthePVPtheory,itisamustandprerequisitetomakeapre·studyonthecorenotionof‘valency’．2．3．1DefinitionofValencyAttheendof19thcentury,MaJianzhongdiscoveredthestructureandmeaningconnectionbetweenverbsanditssubordinatingnounsinhisMaShiWen而馏．LvShuxiangalsodiscussedvalencenumberofChineseverbsinhiswork,whichwastheembryoofverbvalency．Inthework“OntheWordDe，Destructureandjudgmentsentence”，ZhuDexi(1978)firstlybroughttheexplicitconcept“valencygrammar'’intoChinesegrammar．Laterin“Self-referenceandTransferred-reference”，heinnovativelyintroducedtheconcepts“syntacticgap’’and“Extraction”．Theseresearchesarousedalotofscholars’intereststowardsvalencygrammar,whichhighly ThesisforMasterDegreeofZhejiangUniversityChapterTwoLiteratureReviewpromotedthedevelopmentofChinesevalencygrammar．Afterthat，ZhangBin，WuWeizhang，FanXiao，LiaoQiuzhong，LiuDanqing，YuanYulin，TanJingchun，ZhuJingsong，WangLingling，ShenYangandLuJianmingsuccessivelypublishedaseriesofarticlestoprobeintovalencytheoryandUSethistheorytOanalyzetheChinesegrammarphenomenon，andalltheseunveiledthevalencytheoryandputitintoanotherlevel．YuanYulin’S“StudiesonChineseVerbs’Valence’’(1998)alsobecamethefirstmonographofcontemporaryChinesevalencygrammar．FengWaSgenerallyconsideredasthepioneerbringingdependencygrammarintoChinesecomputationallinguistics．In1980s，headoptedthevalencyconceptwhichwasraisedbyTesni6reintheMT(MachineTranslation)researchandproposedapreliminaryChineseValencySystem．GroundedontheCVS，aMultiple—branchedMultiple-labeledTreeModel(MMT)wasputforward．Littleresonancecouldbeheardatthattime．Tillearly1990s，agroupofTsinghuascholarsheadedbyPro．Huangchallenginglyendeavoredinbuildingupadependencycorpusbasedondependencyrelationstagging．InformationRetrievalLabfromHarbinInstituteofTechnology(HIT-IRLab)sharedaChineseDependencyTreebankandmadegreatcontributioninthisarea．Alsosomeotherdependencytreebank,ChinesePennTreebank(eTa)Sinica．analysisexperimentsaremadeinCKIPandSinieaCorpusofTaiwanAcademiaR．EmonsandD．J．AllertonmadethebiggestcontributiontOvalencygrammarofEnglishlanguage．RudolfEmons’“ValenzgrammatikfardasEnglische’’isdesignedaSauniversity—levelintroductiontovalencytheory．Heexplainedtheterm“valency’’andalsodefinedthat‘theverbdeterminesthenumberofelementsnecessaryforacompletesentencebythefactthatexchangingonlytheverbcanturnawell-formedsentenceintoanon—sentenceandviceversa’．Healsolistedthecomplementcategoriesandsubcategoriesofeachcomplement嬲thenecessaryfactorsforacompletevalencydescription．D．J．Allerton’S(1982)bookonvalencyintheEnglishlanguageisdifferentinmostaspects--audience，scope，methodandaimmfromEmons’．Itdoesn’taimfor19 !!!!!!!竺坠!!堕2矍翌121兰垒旦!竺竖芏竺∑!翌!里g望12坚!12生!塑!哇竺坠∑!!∑coverageasextensiveasEmons’workbutratheramoredetailedandin·depthanalysisofasmallerfragmentofEnglish，namelysimplesentenceswithonlyonelexicalverbandseveralotherrestrictionspertainingtocomplementtypes．What’Smostimportant，AllertonmadeathoroughinvestigationintoEnglishverbsintheperspectiveofvalencygrammar．Inviewoftheimportanceofthevalencytheorywithingloballinguistics，itisratherremarkablethatitsapplicationtoEnglishhasbeenrelativelylimited．PerhapsitissignificantthatthefirststudiesdevotedtotheanalysisofEnglishcomplementationwithinavalencyframework(Emons，1974and1978；Allerton，1982；HerbsL1983)werewrittenatEuropean(inthesenseofnon-British)universities，howevertheconceptofvalencyhasbeengaininggroundinBritishlinguisticsinrecentyearsaswell(Leech，1981；Matthews，1981；Somers，1984)，especiallysincescepticismtowardsgenerativelinguisticshasbeengrowing．Furthermore，aconsiderablenumberofapproachesnotdirectlylinkedtovalencytheoryhavepursuedideassimilartothoseofthevalencyapproachinrecentyearsbyputtingmoreemphasisonthecentralroleoftheverbetc．Nevertheless，acertainneglectofvalencytheoryamongstBritish(andespeciallyAmerican)linguisticscannotbedenied．ThismaybeduetothereasonthattheoriesnotoriginatingfromEnglishmayhavereceivedlessattentionintheEnglish-speakingworld．Afterexpressinghisworriestowardsthevalencytheory,T．Herbst(1999)proposedtodesignaEnglishValencyDictionary．Eventuallyin2004，hehadpublishedthebookAValencyDictionaryofEnglish，wherehedidaprofoundinvestigationintoEnglishverbs，nounsandadjectives．2．3．2PVPTheoryanditsRelatedResearchIn2007，FengZhiweiandLiuHaitaoproposedaninnovativeconcept“ProbabilisticValencyPatternTheory(PVP)”．Onthebasisoftraditionalvalencytheoryanddependencygrammar,aprobabilisticandstatisticalfactorwasconsideredintothisnewPVPtheory．Itwouldanalyzethegrammaticalandsyntacticalfunctionsofwordsindualperspectives：quantitativeandqualitative．Thecohesion(valency)ofwordisconstitutedbycentripetalforce(input)and ThesisforMasterDegreeofZhejiangUniversityChapterTwoLitexatureReviewcentrifugalforce(output)，inwhichtheformerindicatesthepowerofonecategorygovernedbyothercategories，thelatterindicatesthatthepowerofonecategorygovernstheothercategories．TheypointedoutinPVPthatqualitativeresearchcoulddescribethatthedependencyrelationscouldgovemandbegovernedbyonewordcategory,alsoquantkativeresearchwouldhelptospecifytheproportionandfrequencydistributionofthesedependencyrelations．Figure2．7coulddescribethefundamentalprinciplesofPVPtheory：GlGz⋯⋯G胁lGhWDlDzD3⋯Dm2D协1DnFigure2．7DiagramofPVPTheoryHere，“W”coulddenoteanyspecificwordoranypartofspeechinthesentence．“Gl，G2⋯Gn”denoteallthedependencyrelationswhichgovern“W”，while“D1，D2．．．Dn’’denoteallthedependencyrelationswhich‘‘W’’govems；‘‘wgl，w92．．．wgn’’denotetheprobability(weight)ofallthedependencyrelmionsthatgovern‘‘W’’respectively,while‘ⅧI，wd2⋯wdn”denotetheprobabilityofallthedependencyrelmionsthat“W”governsrespectively．Asaresult，aformulacouldbeobtained：wgl+w92+⋯+wg．=1，wdl+wd2+⋯+wd．=1．Inthisway,thedistributionofdependencyfunctionofonespecificpartofspeechcouldbeauthenticatedanddisplayedexplicitly．LiuandHuang(2006)proposedContemporaryChineseDependencyGrammar,whichcontains24partsofspeechtagsand34dependencyrelmionstags，andthedependencytagsetincludes20complementsand14adjuncts．TheChinesedependencytypessetbyLiuandHuanghavebeenlistedontheAppendixII．Onthebasisofthisgrammar,FengandLiu(2007)conducedasurveyonsortiepartsofspeechwithgreat2l ThesisforMasterDegreeofZhoiangUniversityChapterTwoLiteratureReviewcohesion，suchaSverb，adjective，noun；diagramsofinputandoutputareonbehalfofthegoverningrelationsanddependingrelationsseparately．ThreeexamplesarelistedinFigure2．8，thesethreegraphsstandfortheweight／probabilityofdependencyrelationsformedbyverbsaSgovernor,dependent,nounsasdependentindividually(Thebreadthoflinessymbolizedtheprobabiiityofthedependencyrelations)： ThesisforMasterDegreeofZhcjiangUniversityChapterTwoLiteratureReviewFigure2．8DependencyRelationsRepresentationsofChineseVerbandNouninPVPBuiRontheconceptsofdependency订eebankandPVP,GaoSongfromCommunicationUniversityofChina(cuc)extendedaresearchonChinesenouns．Sheselectedtranscriptsfrom31programsincludingtelevisionshowsandbroadcastingnews，whichcouldensurethediversegenresofcorpus．Thewholecorpusconsistsof3600sentences，98236words．SheemployedtheprofessionalPOStaggingsystemsdevelopedbyChinaCommunicationUniversitytodotheautomaticwordsegmentationandpart·of-speechtagging，andthenmanuallyfinishedthesyntactictaggingintheviewofdependencygrammar．StatisticalanalysisofExcelWasthenappliedtoclarifytherelationshipsamongthegovernorproperty,subordinatepropertyanddependencyforms．HerexperimentalresultsconcludedthedependencyrelationsformedwhenChinesenounsgoverningcomplementsandexplications，andthegrammaticalfunctionswhenChinesenounsplayasthegovernorandsubordinate，andthensheobserved6grammaticalfunctionsofChinesenounsincludingsometypicalandatypicalfunctions．Itisarelativelyup-to-datetheorywhichWaSraisedin2007byFengandLiu．Previousdependencygrammarresearcheswhichhadbeeninvolvedthestatisticcomputingwereonlyappliedinthenaturallanguageprocessing，butnotrealizedintheoreticallinguisticsdomain．Atpresent,theinnovativePvPconcerningthequalitativeresearchhasonlybeenemployedbyGaoSonginChinesestudy．NobodyhasbeenfoundtousePVPtheoryinEnglishdependencygrammarresearchyet．Figure 里堡!!!塑塑兰!塑望!!!望21兰垒复i竺墨旦尘：!!!堡妄垫!旦坚!塑兰堕竺!!竖坠∑!!12．9presentsthepartialresultsfromGaoSong’Sresearch．支配荚系频次比例倒旬宠疆a酽l1201873．70％薅曲l}．勒鲥毋：村飘：：t强{：太磊．}+j錾结炭0。妊点符号p眦萨327520．og％翔袅珐“l迓上丝．哥丘：淮赫＆住总瑗，H纯漆epa3312．∞％母j。j{‘0挂¨笈表r蘑《漭滋。J：剃天系COOl"1270．78％，串爱暖务笺砭!加钕臼、!j苑品奄套妊嚣雏援；0。拨诱adva12lO．74％饱矗：龟税l：黼淡沦盟二尘∞Ⅲ』翘。恐碗戮荔芫系a假r1110．68％役多彦≈．久鬣盘fjt，私儿爷警多种爱嚣．癌。复钓笑系cf108O．鼯％今天互i!丛．缝型怂垓好jf歇歇j’。￡话蜘bj弱0．34％址小过jj；耷麴监，名列复数pIc50O．3l％我为我冒航天率盥傲越贞缴鹩互二三剑敛铱。匈泰霹嬲语e鞠37O．23％这是纷么业?鲢翊补是话qc33O．20％该颂j：程店动j，二兰多．成效爱著。迄带艽系e∥3lO．19％弱竺露耍眈z：显军陵．五星{J效：蠹盈。摇入逶m80．oS％兰出芷。像麓辐豹玉互．套i{16306loo％Figure2．9FindingofChineseNounsastheGovernorinGaoSong’sStudy ThesisforMasterDegreeofZhejiangUniversityChapterThreeResearchMethodologyWiththepurposeofrevealingtheveilofdependencyfunctionsofEnglishpansofspeech，anEnglishdependencytreebankhasbeenconstructedtofigureoutwhatroleEnglishnounsplayinthedependencyrelationsthroughadirectstatisticalanalysis．Thischapterreportsthemethodologyappliedinthisresearch．Researchquestions，materials，instrumentsandresearchprocedureswillbeintroducedindetail．3．1ResearchQuestionsThisresearchattemptstoanswerthefollowingfourquestions：1．ThroughwhichdependencyrelationswouldmaketheEnglishnounsgovernotherwordclasses?Whatistheproportionoftheserelationsandhowtheydistribute?2．ThroughwhichdependencymlationswouldmaketheEnglishnounsbegovernedbyotherwordclasses?Whatistheproportionoftheserelationsandhowtheydistribute?3．Aiieranalyzingtheabovetwoquestions，whattypicalgrammaticalfunctionsofEnglishnounscouldbesummarizedandwhatatypicalgrammaticalfunctionscouldbecomplemented?4．ComparedwiththeChinesenounsresearch，whatarethesimilaritiesanddifferencesofChinesenounsandEnglishnouns?Theanswerstothesethreequestionscanbeformulatedinlaterfindings．3．2MaterialsAnEnglishdependency仃eebankwillbeusedinthisresearch．Toguaranteetheaccuracyandfeasibilityoftheresearch，thetranscriptsofNewConceptEnglishandAdvancewithEnglishareselected嬲therawmaterialstobuildthecorpus．whichcomprises4volumesofNewConceptEnglishand1modulesofAdvancewithEnglish，precisely,6280sentencesand101822words(includingpunctuation)．Thevariousthemesofthesetwobooksinvolveaspectsofsociallifeandscientificdevelopment，andalsoreceiveagreatdealofpositivefeedbackfromthecirclesofstudentsandteachers．Meantime，thelocallanguageusageandappropriatedifficultylevelvalidatethesignificanceofadoption．Figure3．1displaysthescreenshotofthis transcript．鍪唆墼篓：鲤经罗嗍趁嬲舅蹴琵瀚疆———■凌隧搦凌麓琵豹兹缓戮滋■■翻瞄嘲—————■——■瞄糍鎏尊口锄靴毪’i一鸷≈’‘抵‰飞⋯⋯⋯⋯⋯。⋯。⋯⋯⋯。⋯⋯。。⋯??～。⋯。。。⋯Don'1takcachance,dthyom-孵ddlnspi∞tosraphs．1li曲口tIalityr．fofesel∞mwedclir镕曲。to盯a舾OcemCc吐re2丁47l鼹8Bvervbride-alns}mtwed出ngtobetheDtost-ond盯f、￡dsrofher1if乞Butp盯先ctueddinglreqairec==efulplmrtlL_}】ethrthebrideandbride嗣rotm·弛pl柚ni珥·tnditimml靠dd抽●∞j‘埘-reldwu吖。娥钾me．P^n唧forthe孵ddins#hicbis札-j”aproble机耐llpartlydeci电theclwlee-Thebzide"-clothes，Wnlch口eusuellyeIp∞ive．requirecm-efulm!lectimIjfⅣbrideschoosetoa_c缸a以te-eddingdreoof口oregistz3tc竹口啊l≈‘舶d●rtda畦n晴-潮I葛●—髓细●tl’点t“-ld∞嘲畦n‘bw,quet．Both口-eco#flytobuybutthEte缸·e幽op#fira血一nicht}l掣canbehirdThesec叽al如S∞plytheFo曲’日outfitn：adim廿j．ckctfarthe唧t矗瓢髓竹-哪～Indtl'五ti∞^ln妇f蕾thehn口睇kFlowers皿‘elmportmtin哪t-eddir蟮；．n碡bride瓤dthebride啪id喇∞m吐ehaBdnomtnmallyc棚=Tyb蚴aezs．CentuB-iesago．*e$tcrnbridesc重Tledh勘c}培；offla-盯s，‘懈ofWaieb-c研veryh蝴-'-tllil‘toI暗砷e—I·，irit·—·_非N啕ad砷喀．couple=prefer■-缸efira盯∞tna-erg．iothey妇1dc∞t扯t4‘∞dflorl=tThe-eddinsStatiqⅥ口y暖亨includeim,ita：ions,phlrtoelh珥，眦巩etc．7heco蛳leneedtoc∞州taspeciali_tprim雷．=hOcan5iveb口lp“advice．Ia哪peopletravelto4-eddiI堪byc茁．Itwouldbeterribleifthec竹brokedean∞the-玎tothecerelmr班,蜘thecoupleelwuldrem—bertohk-州■证melieblecn^fterthewed(finmthebrideand口o∞n奶rhueatradltl删bzm舢et皿’0-e戤目?nlt—ereceptioBTheyneedtore00mtherest日F可Haf记selectthe衄u■∞U俘beforethe-ed以rI‘People咖raalttobediff盯elTtcouldholdthe州酬in4p址叮∞abont．Ye=rsaftermcds,the衄riedct∞le-i11-∞ttolookbeckattheirweddingphotos．Theyshouldresist*hetemptationTOa；kafm,ou—tcuncleto'吐ethepbmocrapb丑№唧begood_ithhiopocketmato=etic．buthewa1曲roe'gettoputin5roltoffil■1ti8hettertobookaphot口舡’am盯_|∞sptci札i∞sin-e托n‘5Thec口t舻lethnuldal‘othink吐．口uttneplaceI南erethepho’oF雄herwilltakethephotos．Itdoesnotelw剞'shzl^，etobe∞theFtepsofthere|ifftrFoffice．Forcz卸h‰anccoupledecidedto}I_pctheirstakmina竹i血n‘pooll1ntheseezeTclseyou_illhe=ra删ti吼follo-edbyastntm=宣lt叮小时埘哪州i血Lookettheinform．tion■rovlded．耐thefour∞sBibleaPJwei'$．■hichpackageofp}-叶alr印hicse门一cesdoesthe-椰-∞-田什tab●，t缸he=-cd出．zI口胁eyoudecided出眦tthephotos妇．m口,,eddinSyr．，I舯h11．DdIreallvlikeavideo-butit’t50ezpe雌ive．r11t吐ee缸eofthat．I知t-0叶everythinstoberi‘lItfor，嘲．‰11．thenI。dliketMs叩己If咖to盯茸t'艘-scametothe-e‘吐珥md弛oeptto=．=n4·喀0■tp酿∞—Idbslf时Vi如．Thatwouldbc-口nd盯“．-michprr嘲tdoesthe-■■n如■efor}l耳frlmd"·-|d击珥7Cml'蟹lpy口u．_目d耐Yes,ple￡me．1‘d1iketochoiceaweddinlpreselltfor_frlend"t-ed她Hern嘶’5IIn∞,ltopkirni．Ibelieve，a皿。or0Idzinsher-edd崦lin．yes,thm’sri曲t．I’11j∞tgetitf=ryou．陆e．∞5tthi珥5hd他缸readybeenchosen畸theot}-贯■翊n^these●畔the嘲thethm,el盱‘I袱．■c11．I’dliketogiveb0啦cutl哪re5lly-B眦IcartEiveh盯kniveLSoI’11givethesesilvera糯4i*。g自H毽％；；I；自；～|2j‘。二：：j+：，名：一j?ij：。：。2：“÷ij÷鼍fz‘二!。■‘拿竹?⋯’‘““，“|博’皇：?j争；ijjjj”。：j叶’“E。。jc_，‘t3．3InstrumentsFigure3．1ScreenShotoftheTranscriptTherearctwoprincipalinstrumentsintheresearch．OneistheStanfordDependencyParser；theotheristhesoftwareExcel．Theformerwillhelptofinishtheautomaticpart-of-speechtagsanddependencyrepresentationslabels；thelatterwillfacilitatetoattaintheultimateresultsofthisresearch．Besides，somecomputerprogrammingtechniqueswillalsobeusedhere，forinstance，thetextfileoftranscriptwillbeautomaticallygeneratedintoExcelfile．AexamplegeneratedfromthesentenceintoExcelfilecanbeobservedinFigure3．2．1’句子编号甸中词序词2；1；1⋯⋯一“”●”’“’p^“‘”^_‘。⋯⋯⋯‘一。—副．⋯⋯⋯工j⋯⋯⋯—。24{1；3⋯—“吨⋯“≯⋯—墨立⋯一k——⋯叁ltrs．Smith’skitchen5is6smallFigure3：2Auto-GeneratingExcelFileoftheSemence：Mrs．Smith’skitchenissmall． —Th—es—is—for—M—ast—er—De—gr—ee—of—Zh—ej—ian—g—Un—iv—ersi—tyChapterThreeResearchMethodology一3．3．1ParserIncomputationallinguistics，parsing(formally,syntacticanalysis)signmesthetext(asequenceofwords)analyzingprocesstodetermineitsgrammaticalstructu旭withrespecttoagivengrammar,forinstance，phrasestructure野lmm盯觚ddependencygrammar．Accordingtocomputerscience，aparserISoneotthecomponentsinaninterpreterorcomplier,whichCallcheckthesyntaxandbuildadatastructure．TheparseroftenUSeSaseparatelexicalanalyzertocreatetokens缸Dmthesequenceofinputcharacters．TheyCallbeprogrammedbyh锄dorfsemi-)automaticallygeneratedbyatoolbyprogramminglanguages·Inthesystemsofmachinetranslationandnaturallanguageprocessing，hum觚l锄guages眦usuallyparsedbycomputerprograms．Toparsenaturallanguagedata，researchersshouldfirstlyadoptonespecificgrammar．Thelinguisticandcomputatiohalconcemswillinfluencethechoiceofsyntax,forexample，SOreeparsmgSvstemsapplylexicalfunctionalgrammar,someotherpopularparsingSystemswouldusenlehead．drivenphrasestructuregrammar勰linguisticformalism，andanotherpopularS仃ateg)，foravoidingcontroversyintheparsingcommunityISdependencYgr锄marparSmg·一Atleast．mostmodemparsersarepartlystatistical，fortheyrelyonacorpusoftrainingdatawhichhasbeenannotatedmanuallyalready,throughthisway,thesystemcouldbeallowedtogatherinformationaboutthefrequencyofonecertainconstru曲oninthespecificcontexts，andthemethodhasbeenwidelyacceptedasthemachmele锄ing．Parserscouldbewrittenbyhandorgeneratedbyparsergenerators·rheresearcherswouldinputprogramminglanguagestocreatesomeformofIntemalrepresentationsforparserswhichwouldinvolvetheparsingalgorithms·ThishasalreadvbecomeapopularandmaturetechnologyincomputationallingulStlcs‘nefollowingexampledepictsacaseofparsingacomputerlanguagewltn铆olevels0fgrammarinvolved：lexicalandsyntactic，whichCallbeseeninFigure3·3： ThesisforMasterDegreeofZh6iangUniversityChapterThreeResearchMethodologyC一～]il篙篡笔鬻l}C～]l僦能∥·ll≮≯E一～]|吲‰黔l(～]Figure3．3ProcessofParsingaSourceString(ComputerLanguage)3．3·1·1TypesofParserHowdoesaparsertransformtheinputintotheexpectantoutput?Generally,theparserwillfinishthistaskintwoways：top-downparsingandbottom—upparsing．Top··downparsingcallbeviewedasanattempttofindleft--mostderivationsofallinput-streambysearchingforparsetreesusingatop-downexpansionofthegivenformalgrammarrules．Tokensareresolvedfromlefttoright．Thecommonparsersfortop··downparsingcontainLLparsersandrecursive--descentparser．Bottom-upparsingcanstartwithinputandtrytorewriteitintothestartsymbol，whichattemptstolocatethemostbasicelements，thentheelementscontainingthesebasicelements，etc．TypicalexamplesforthiskindofparserincludeLRparsersandShift-Reduceparsing，CYKparser．Therearealsotwoessentialdataannotationtypes：constituency(phrasestructure) ～‘“一ThesisforMasterDegreeofZhejiangUniversityChapterlhreeResearchMethodologyparsesanddependencyparsers．In1965，Oaifrnanprovedthatdependencygrammarandconstituencygrammararestronglyequivalent．However，aroundtheearlieryearsof21砒century,anumberofTreebank-trainedstatisticalparser[Collins(Collins，1999)，Charniak(Charniak,2000)，Stanford(KleinandManning，2003)】werecreatedtogenerateparseswithhighaccuracy．netreebanksofthattime．particularlythePennTreebankcouldonlyprovidephrasestructuretrees，andhencethisistheoriginaloutputformatoftheseparsers．Tomeettheincreasinginterestanddemandindependencyrepresentationinrecentyears，somedependencyparsersaredevelopedunderthehelpoftheestablishmentofdependency仃eebanks．3．3．1．2StanfordDependencyParserAmongtheearly-stagedependencyparsersforEnglish，suchasMinipar(Lin，1998)andtheLinkParser(SleatorandTemperley,1993)．theyarenot邯accurateandrobustasthoseconstituencyparsersbuiltonalargecorpora．Tofillthegap，agroupofscholarholdinghighaspirationindulgedintothetaskofgeneratingadependencyparser,whichcomprisesonescholarfromLouvainCatholicUniversityofBelgiumandtwoscholarsfromStanfordUniversityofUSA．Marie-CatherinedeMarneffe，BillMacCartneyandChristopherD．Manning(2006)designedasystemforextractingtypeddependencyparses(trees)ofEnglishsentencesfromphrasestructureparses，whichiSalsocalledasStanfordParser．ThissoRwarecouldbedownloadedonthewebsiteoftheStanfordNaturalLanguageProcessingGroup．。ThedefinitionsonthegrammaticalrelationssetbyCarroll(1999)andKing(2003)havebeenadoptedinthissystem．‘111egrammaticalrelationsalearrangedinahierarchy,rootedwiththedependent．WhentherelationbetweenaheadanditsdependentCanbeidentifiedmoreprecisely,relationsfurtherdowninthehierarchycanbeused．’TheoverallhierarchyofthegrammaticalrelationsisprovidedinFigure3．4，andspecificexampleforeachrelationcouldbefoundinAppendixI．Thehierarchycontains48grammaticalrelationsintotal．ManyNP-internalrelationshavebeenborrowed，besidesthecommonestgrammaticalrelationsforNPs，someothergrammaticalrelationshavebeenadded，includingappos(appositivemodifieO，nn(nouncompound)，hum(numericmodifier)，number(elementof ThesisforMasterDegreeofZhejiangUniversityChapterThreeResearchMethodologycompoundnumber)andabbrev(abbreviation)．卸一姆rldellt雎鼍-lmxiliaty-ll鼍P曩警5+p嘲暗彝蟹■腻蠊腑l了cop—cDF“lIco嘲一∞螂掰l髓霞一000fdma垅ion翻噻’目rgument辩蛹一虢内泌戤嬲捌一nomimfl霸—妇l鞘赫颐p毡器嚣一passive捌群d桶l嚣—，萤蜮c鬟甜巧一d砌l嚣鑫|飘lb参缎啪咿’a獬k黼t嘲一。锄。cld。扫i—directobk-etiobj—mdinzet曲麟蠼弘慵啄一萌琢短t耐p哐q∞赫峨哦搬一删uribmiveccorv,p·eb嚣鬟甜oD靠坏墙目啊嘣twithil嘲f删确jeetxc站mp·dausai∞mp醣舶}l鲨铆i妇阻薅．犍糟砒懿lb画c穗c掰疆蝉一戗珊甲妊’，耘mt酾∞聪I置rk一露堪“嚣rlw槐xl妇哟屯硅晒，壤麟翻劂)掰l-relativef铺艇肾d函m：,dacmg量rcmod)_掰棚妒一adrenal盘搠Bp醅l目c礅l妻霉暾‘丑{够豫辩f．refcmm麟堪-ezpietiwe缸蕾p董幽雠旃删rood—modifmr叠‘蛔d．arise'hidan翰}modifierp啪1Kl一弘嚣阿獭ccl擞l鹫rood／fret船韵昭d—tmporalmo蕴ffg-rn嚣n喇，I棼王|‘6vcd蕴E‘s譬匏戴x舡lic旷amod一划叠戤k越modifierinfmod．iafinhiv越mo碰fiervaru'洲一panici|，i童lmod／fa口棚《嚣．霸勰#￡鬻：6emodlf杞r秘堪嚣Ib凹-clc驺鬻滋￡ofe口啊Dot狮d口瞳u撕a，巴fI善串％·墨疆x’藿i噍翊霉lmodifier舶一no'dinGo枯朝麓M圩猫modifier_扫b∞F—ab￡就vh吐io科盯渊i鲔嚣“●曲靠剃．叠dverbialn渊i《；材neg—negja．ion曩_。吐涵饼F峙繇一pE强孵s蕊∞moaifiefp馥堪岛钳Ii犍一po释勰孳菇雠翻鼢dj西防t：，》尹一·萨lr＆#鲥w礁’p_蒯cb如t．d学k：l。￡l矗nof争∞p-孵p∞溺槐l嫩搠磁磊嚣r赋kp一舅}舶嚣珏6e姆翱堑圈膳勰嘶一崩撤∞mIil绣％调哆啾Figure3．4GrammaticalRelationHierarchyThetechniqueforproducingdependencyparsesisprimarilybasedontherulesusedonphrasestructuretrees．Themethodhastwophases：dependencyextractionanddependencytyping．Inthefirststep，asentenceisparsedwithaphrasestructureparser'andahighaccuracystatisticalphrasestructureparser(KleinandManning，2003)trainedonthePennWallStreetJournalTr．eebankiSusedhere．Theheadofeachconstituentofthesentenceisidentified，usingrulestoretrievethesemanticheadofthe ThesisforMasterDegreeofZhejiangUniversityChapterThrceResearchMethodologyconstituentratherthanthesyntactichead，whichiscrucialforextractingsemanticallyappropriatedependencies．Determiningheadswillenabletofindthecorrectdependencies．Inthesecondphase，eachofthedependenciesextractedwithagrammaticalrelationwillbelabeled．Foreachgrammaticalrelation，oneormorepatternsOVal"thephrasestructuretreeswillbedefined．Theneachpatternismatchedagainsteverytreenode，andthematchingpatternwiththemostspecificgrammaticalrelationistakenasthetypeofdependency．ComparedwithotherdependencyparserslikeMiniparandtheLinkParser,theadvantagesofStanfordDependencyParsercouldbeviewedinmanyaspects，especiallywithrespectstoaccuracyanddependencyrepresentation．Aper-dependencyaccuracyof80．3％couldbeobtainedthroughthetestofsamplesize．TheStanfordDependencyParseranddependencytypessetinthissystemhavebeenappliedinmanyresearches，andalsohavewonthehighreputationaroundtheworld．SomeresultsautomaticallygeneratedfromStanfordDependencyParsercanbeobservedinFigure3．5．⋯1。～哇kitchenliN，⋯．smallJJ’nsubJ：⋯王。5is．gBZ：6⋯smallJJ⋯cop⋯。1⋯，6small，JJ．0，。⋯⋯j⋯17．．0’⋯”4一”⋯，21ThereEx’2isVBZ’crpl2is船Z023aDTj41refriger￡脯det⋯⋯2⋯⋯⋯～4refrigerat肼⋯⋯j2⋯⋯⋯is⋯⋯⋯昭三⋯⋯nsubj⋯～25inIH026theDT’7kitchen舳det27kitchert职j哇refrigm：eSliprepin23甚-⋯_⋯⋯⋯，r～。一——9～．⋯⋯⋯，⋯⋯1Thefir’2‘re士riger掮Nde％2refrlgerat脯≥⋯twhiteJJnsubj3is⋯⋯VBZ～1⋯⋯：white，JJ；cop哇white⋯⋯⋯JJ⋯⋯⋯0⋯⋯～⋯⋯⋯⋯⋯5，．．0Figure3．5ResultsofStanfordDependencyParserinExcelFormat3．3．2ExcelThestatisticalfunctionofMicrosoRExcel2007willbeusedtoanalyzethe3l !!竺!虫!坚呈苎竺21墼121圣堕!竺兰望坐竺型竺翌型!!旦坚壁!!磐!坚业2塑12墅frequencyofdependencyrelationsinwhichnounsactastheheadandthedependentrespectively．Andthissoftwarewillautomaticallyrankthedependencyrelationsaccordingtothefrequencyandproportion．AfractionofthedirectstatisticalresultsfromExcelhavebeensortedonFigure3．6．“?■二j‘_≯一?4=‘∞q目_■■“目__■_目∞∞__—々“自_《__m鹇““∞Hw__捌自∞_《∞⋯一—!“_曲啪∞％t～⋯—”肖鼎“_∞∞_Ⅲ曲㈧—““_mⅫⅢ∞_Ⅲ制∞_m_黼Ⅻ却“‰一名词支配依存关系⋯次数比仞名词从属依存关系∞渤比例⋯⋯～嗽⋯⋯⋯⋯⋯⋯一8828⋯。．簟．52％⋯⋯mub、⋯⋯⋯⋯⋯⋯⋯nod468817．2缁⋯⋯由hj⋯i⋯～m⋯⋯⋯⋯⋯。⋯⋯⋯～2233⋯龟．礅——⋯～m⋯⋯⋯⋯⋯～⋯⋯⋯～poss⋯⋯⋯1820⋯6．70％prep．of。⋯⋯．疆％奠⋯⋯⋯⋯——⋯1637⋯⋯⋯龟．03W⋯⋯。pr印in⋯⋯⋯⋯⋯⋯⋯⋯。：⋯⋯。conj—aIId⋯⋯⋯⋯⋯⋯⋯⋯7驰⋯⋯～2．94％⋯⋯⋯⋯conj．and⋯⋯⋯⋯⋯髓．。⋯7742．嗽⋯⋯⋯⋯～pr印．to⋯⋯⋯⋯⋯⋯。⋯～mubJ～⋯⋯⋯⋯⋯759⋯⋯2．啷⋯⋯⋯～却⋯⋯⋯⋯⋯⋯⋯。?⋯娜⋯⋯⋯⋯⋯、蛆⋯2．75％⋯⋯，璐逝X瞰s⋯⋯⋯⋯⋯⋯⋯；，．。。一。rcaod⋯——715一——2．63％～⋯⋯～prep_for⋯⋯⋯，⋯一—。⋯⋯⋯⋯dep，6tl2．姗⋯⋯。，皿印-∞⋯⋯一⋯⋯。．——一prep_in～⋯⋯～⋯一⋯⋯，lib⋯i．撼⋯⋯⋯⋯⋯prcp．dth⋯⋯⋯⋯⋯⋯⋯⋯一。：．。advlod⋯一，⋯4331．60％prep_at．⋯⋯。⋯⋯娶缸、础————～⋯衡⋯⋯～t。05％⋯⋯～⋯tsod⋯⋯⋯⋯⋯～⋯⋯，——⋯；⋯⋯～a1]pos～⋯⋯⋯⋯⋯⋯⋯⋯，266⋯0．9麟⋯⋯⋯a挪∞一⋯～⋯⋯。⋯．皿印jor2190．B壕⋯⋯．poss⋯⋯⋯⋯⋯⋯⋯一⋯～conj—or⋯⋯⋯⋯⋯⋯⋯⋯。164⋯⋯⋯0。60％。F印-frn⋯inf-od⋯⋯⋯～憾⋯⋯0．58％⋯⋯⋯pobJ⋯⋯⋯⋯⋯⋯⋯～j⋯⋯⋯pro_on⋯⋯⋯⋯⋯⋯⋯⋯～138⋯⋯。0．5暇conj．or⋯⋯⋯■auI⋯⋯。1220．憾⋯⋯⋯⋯prepinto⋯⋯⋯⋯⋯⋯⋯⋯～?o⋯We呶⋯——⋯一⋯⋯⋯m—～0．谌⋯agmt⋯：⋯⋯一pr印_to～⋯⋯．83⋯0．31％⋯⋯⋯⋯⋯prepu⋯⋯～⋯⋯⋯⋯⋯o～⋯pr印vith⋯⋯～82⋯0．3嘣⋯⋯⋯prep_about⋯⋯⋯⋯～ladvcl⋯。720．27％⋯⋯．cco|p一⋯。。——。．————，⋯～neg⋯⋯⋯⋯⋯。⋯⋯⋯⋯镐⋯?0．25％⋯。pr印_by⋯⋯⋯二⋯。prcpc．．of⋯⋯⋯⋯⋯，～?68⋯⋯0．25％⋯～x嘲⋯⋯．⋯⋯⋯COlpll。⋯⋯62⋯0．23％．．～F印_d咐⋯⋯⋯⋯～⋯⋯⋯prep_a!⋯⋯⋯⋯⋯⋯⋯⋯。。60⋯⋯⋯0．22jc⋯⋯属印_lik!——⋯37蛎18．2器36∞IL5鼎222310．8揣18719．1荔12766．2荔晒0．3荔Figure3．6StatisticalResultofExcel3．4ResearchProceduresThisresearchwouldbeimplementedaccordingtothefollowing6procedures：3．4．1ConstructingtheCorpusOver6000EnglishsentencesfromNewConceptEnglishandAdvancewithEnglishwillbeselectedasthematerials；andallthesesentencesinelectroniceditionaledownloadedorboughtfromthewebsite．Tofacilitatethesyntacticstructureanalysis，sometranscriptionsoftheoriginalmaterialhavebeendonebyhandwhenthe帆；骞饼镰噶螂儒慨蹴{鸯獬姒Ⅲ嘣嗽慨倦嗡豫镰弧l§毛＆ZZ互ZLL仉以仉仉m乱仉仉仉l曼m啪啦m雠善；嚣；啪撇撇渤猫m懈煅扭啪懈啪∞一他 里!坐鱼!坚竺塑里!受121圣堕!竺墨业：里盟￡垒苎坐!：!垒竺!垦!!!竺!坠!生2塑12墅sentencesalebeinginputtedintonotepad(txtversion)，suchasdeletingsomeexcessivepunctuation，transferringthedirectspeechintoindirectspeech，removingtheunidentifiablecharacters，andSOon．AccordingtothestatigicalreportofExcel，thiscorpusincludes6280sentencesand101822words，includingthepunctuation．Thesearetherawmaterialswaitingfortheprocessingtobecometheannotatedcorpora(treebank)whichcouldbeputintopracticeimmediately．3．4．2AutomaticPart-of-SpeechTaggingAstheessentialpacetoconstructatreebank,annotatingthepart-of-speechforeachsentenceshouldbedonebythesoftwareinthisphase．AgeneralPOStagger(Part-of-SpeechTagger)oftreebankhasbeenembodiedbytheStanfordParser,whichsharesthesametagsdescriptionwithPennTreebank．Thealphabeticallistofpart-ofspeechtagscallbefoundinAppendixIIIindetail．ThesetagswillbeautomaticallyfilledintothespecifiedlistofExcelsheetafteraddingtheprogramminglanguageintotheStanfordParser．3．4．3AutomaticDependencyRelationsTaggingStanfordParserisalsoinchargeofannotatingthedependencyrelationsoftheEnglishsentence．ThesothareisdownloadedfromthepublicwebsiteofStanfordNaturalLanguageProcessingGroup，whoselatestversion1．6．5willbeusedintheresearch。StanfordDependencyParser,abranchofStanfordParser,willautomaticallygeneratethespecificdependencyrepresentationsforthisresearch，bydenotingthedependent,part—of-speechofdependentandtherelativedependencymlations，etc．ThedependencyinformationabovewillbeinputintothecorrespondinglistofExcelsheetincommonwithPOStags．3．4．4ManualDependencyRelationsCheckingDespitethefactthattheannotatingtaskofdependencyrelationsisgettingmucheasierwiththeassistanceofdependencyparser,humansentencesarenoteasilydigestedandparsedbycomputerprograms，forinstance，theambiguityinthesemanticstructureofhumanlanguage，thiskindofcasescouldbecommonlyfoundinChineseresearch，likethevarietiesofphrases．Hence，theresultsgeneratedbythedependencyparsershouldbeexaminedbytheresearchersinascrupulousway．Theinaccurate33 ThesisforMasterDegreeofZheii锄2UniversityChapterThreeResearchMethodologyannotationwillreceiveapromptcorrectioninordertOguaranteethevalidityofthisdependencytreebank．3．4．5StatisticaIAnaly’sisInthefinalstep，MicrosoftExcel2007willbeusedtocounttheljrequency(weight)ofthesedependencyrelationsindividually．Duringthisprocess，Englishnounswillbefilteredwiththepart—of-speechlabelstartingwith‘‘N”(DetailscanbeobtainedinAppendixIII)．Anotherdistinctionwillbemadeamongthedependencyrelationswherenounsfunction硒headordependenLthroughthisdivision，twostatisticalresultswillbeachievedrespectively．AndthetypicalandatypicalgrammaticalfunctionsofEnglishnounswillbeanalyzedviatheiIlustrationofstatisticresults． ChapterFourResultsandDiscussionInthischapter,thedatagatheredfromthepreviousresearchprocedureswillbeanalyzedindifferentaspects．TosolvethefourresearchquestionsposedinThree，findingsdrawnfromthestatisticalanalysiswillbepresentedsuccessively．4．1DependencyRelationsofEnglishNounsastheGovernorThefirstresearchquestionissettodescribethedependencyrelationswhenEnglishnounsfunctionasthegovernorinthesentence，andtoranktheprobabilityofthesedependencyrelationsunderthiscase，sothatthecommonandraredependencyrelmionsproducedbynounswillbedistinctivelyobservedinthetable．Atthefirststep，thehorizontallinewhose“dependingpartofspeech”taggedwiththestartingletter‘'巾’willbeselected，forthegroupof‘‘N”includingNN(singularnounormass)，NNS(pluralnoun)，NNP(singularpropernoun)，NNPS(pluralpropernoun)allpresentEnglishnouns．Then，theinformationfromthecolumnof“dependencyrelations’’willbesortedbystatiSticalfunctionofExcel．4．1．1ProbabilityofDependencyRelationsGovernedbyEnglishNounsThecommonestdependencyrelationsofEnglishnounsplayingasthegovernorinthesentencehavebeensummarized．20outofthewhole48dependencyrelationsCallrepresentallthesyntacticfunctionsperformedbynounsasthegovemonTheoutcomeofstatisticalanalysisinthissectionCallbeexplicitlyobservedinTable4．1．ItshouldbeemphasizedthatthecurrentgrammaticalrelationsusedhereareallextractedfromtheStanfordTypedDependenciesManualwhosedefinitionsarealsoborrowedfromthePennTreebankpart—of-speechtagsandphrasallabels．UnliketheChinesedependencysyntacticdivisionsetbyLiuandHuang(2006)，therulesforEnglishdependencysyntacticrelationsaremadeinamorespecificway,andreceivedunanimousrecognitionintreebankresearchforalongtime．ButsomeoverlappingpartsalsoCanbefoundbetweenthem．35 Table4．1DependencyRelationsDistributionofNounsasGovernorDependencyFrequencyProportionSampleSentencefromtheTreebankRelations(％)determiner882832．98lThearmchairsarenearthetable．det(armchairs，the)；det(table，the)adjectival468817．514Tommyfoundtwosmallcoinsonthemodifierfloor．amod(coins，small)prepositional320111．959Theplanewaslateanddetectivesmodifierwerewaitingattheairportallthemorning．prep．at(airport,at)predeterminer22478．395Alltheschoolboysarecelebratingtheceremony．predet(schoolboys，a11)nouncompound22338．342Ienterforthegardencompetitionmodifiertoo．nn(competition，garden)conjunct9983．728Shemustappearinabrightreddressandlongblackstockings．conj(dress，stockings)numedcmodifier7742．892MayIhavetwotickets，please?num(tickets，two)nominalsubject7592．836Fishingismyfavoritesport．nsubj(sport，Fishing)copula7472．791Itmustbetheonlymodernhouseinthedistrict．cop(house，be)relativeclause7152．671CaptainFawcett’SfirstpassengermodifierwasadoctorwhoflewfromBirminghamtoavillageintheWelshmountain．rcmod(doctor，fiew) 里竺堡鱼!丛竺塑旦!g坚121兰垫堂!竺兰望翌生望!堡￡垒!坐!呈2竺壁!竺!堡竺璺里i!!竺!12翌preconjunct3131．169Bothpartieshavelosttheirreputationincitizen’Sheart．preconj(parties，Both)participal2851．065Asearchpartwentintothecavemodifierhopingtofindburiedtreasure．partmod(cave，hoping)appositional2660．994HarounTazieff,thePolishscientist，modifierhasspenthislifetimestudyingactivevolcanoesanddeepcavesinallpartsoftheworld．appos(Tazieff,scientist)possession2630．983BigBentakesitsnamefromSirmodifierBenjaminHall．poss(name，it)infinitivemodifier1570．587Theonlywaytodothiswastooperate．infmod(way,do)auxiliary1020．38lForhim，thiswouldbeanentirelynewhobby．aux(hobby,would)negationmodifier780．291Thefisharenotsharksorkillerwhales．neg(sharks，not)complementizer620．232HefeelsthathisriseinStatuSiSwellworththelossofmoney．complm(10ss，that)clausal46O．172Thereisapopularbeliefthatthecatscomplementhaveninelives．ecomp(belief,have)abbreviation50．019BritishBroadcastingCorporation(BBC)abbrev(Corporation，BBC)Tbtal26767100ItcanbesgenfromthetableabovethatthestatisticalreportofEnglishnounsasthegovernorinthedependencyrelationsCallfullyrepresentthevarioustypesofnounphrases，andthestrongbindingforceofthenouns．37 Starting舶mthenouns’ownproperty,nounsaremostfrequentlycombinedwiththe如胞册zf加一takingupalmostonethirdofthewholecases．Tobespecific，thedetcnllinersincludethemostcommoncentraldeterminers，thedefiniteandindefinitearticles．andothercentraldeterminers，suchasthis／that／every／each／no，formingasetofclosed．classitemsthataremutuallyexclusivewitheachother．Theinterrogativedete咖iners．wh．group：what,which，whose，thequantitativedeterminers(enough)andassertivedeteminers(some／any)alsoarecontainedinthispart．Besidesthecentralones．therearealsothepredeterminers．Comparedwiththepreviousones，theoccurringfrequencyofthisgroupislower,fortheiroccurringpreconditionistocombinewithcertaincentraldeterminers，andaheadofthem．ne2Ildand7thlargestdependencyrelationsledbyEnglishnounssignifythecombinationofnounswithadjectivesandnumeralstoindicatethequalityandquantityoftheheadnoun．Thepossessio”modifiP，indicatestheheadnounandgenitive’s，whichcomplementsthepossessionofthenouns·Althoughconjunctionsaretreatedasymmetricallyindependencyrelations，conjunctionsconnectthetwoelementsequally，thusthewordsatthetwooppositesidesofacoordinationconjunctionmustsharethesamepart·of-speech．Inthissystem，theheadoftherelationisthefirstconjunct，andotherconjunctionsdependOilitviatheconjrelation．The3一largestdependencyrelationsaretheprepositionalmodifierwhichmeansaprepositionclauseservingtomodifythenoun．Here，prepositionclauseplaysastheadjunctofthenounswithallexplicitindicating．Sameadjunctcanbeobtainedintherelationsofrelativeclausemodifierandinfinitivemodifierwheretheyallcomplementtheheadnouns．Thenounanditsmodifyingprepositionalsomakeupadiversifiedassociation．Themostfrequentprepositionsusedinaprepositional．clausetomodifythenounshavebeenlistedasTable4．2．ThosefrequenciesundertwOhaven’tbeenillustratedinthistable． ThesisforMasterDegreeofZhejiangUniversityChapterFourResultsandDiscussionTable4．2PrepositionModifiersDistributionforEnglishNounsDependencyRelationsFrequencyDependencyRelationsFrequencyprepof937prepbefore11prep—ln776prep_across9prep．to429prep-near8prep_for319prep_accordingto8pr眨on254prepc_togetherwith6prep_with159prep_一ut6prep．at132prep_without6prep_from75prep_towards5prepinto64prep_apartfrom5prep_as51prep_lⅨcept5prep_about48prep_11extto5prep_by36prep_above5prep-0ver32prenjlnce5prep_like25prep_behind5prep_suchas19prep_awayfrom3prep_through14prep_infrontof3prep_under13prep_down3prep_along13prep_alongwith2prep锄ong13prep_upon121’otal3601prep_including12OtherthantheCoordinationinChinesedependencytypesdefinedbyLiu(2006)，thecoordinationheremeanstherelationbetweenanelementofaconjunctandthecoordinatingconjunctionwordoftheconjunct．111isrelationappearsSOrarethatcanbeignored．Apartfromtheadhesionofnounstoformthenounphrase，thesyntacticfunctionperformedbynounscanalsobeobservedfromTable4．1．Amongtheotherdependencyrelations，刀D姗compoundmodifieraccountsfor8．3％ofthewholecases，39 ．ThesisforMasterDegreeofZhejiang—U——n—iv—e—r—si—ty————————————————．．—．．．．C．—h．a，pt—e。r．F—o．u—r．R—e—s—u—lt．s—a—n—d—D——is—c—u．ss—i．o—n●—————————————————，__——．__-——-。●__●____●。__●_______-●___________________________----__-____●___-_____-__●-。_—。_-。_。一．一inwhichthedependentplaysastheattributiveoftheheadnoun．Anotherlargegroupcomesfromthedependencyrelationsofcopula,nominalsubject,auxiliaryandnegationmodifier,andacommongroundcanbereachedinallthesefourrelations，wheretheverbofthesentencemustbeacopula,especiallvforthelaterthreerelations，onlytheverbiscopula,thenthenounscouldplaythegovernor,here，thenounsfunctionasthecomplementofthecopulaverb，alsothepredicateofthesentence．Thetotalofthesefourcasesamountsto6．4％ofthewholedependencvrelgionsunderthiscircumstance．Theoppositesitumioncouldbeseenbetweentheclausalcomplementandthecomplementizer，theformeroneindicatingthatdependentsarethecomplementsofheadnounsinlexicaldomain，whilethelatteronedescribingthenounsasthelexicalcomplementsforthedependents．Aftercalculatingthedatafromthetreebank,aboldassumptionisproposedinhere．AccordingtOthedifferentlexicalrepresentationspresentedbythese20dependencyrelations，theyaredividedintothreegroups：themodifiers，theparallelsandthecomplements．Themodifiersaredescribingthe．relationsbetweenthenounsanditSmodifiers，wherethelattersupplementsorrestrainsthemeaningofthenounSincertaindegree；theparallelsaretherelationsbetweenthenounanditscoordination，sharingthesamepart‘of-speech，whichcouldbereplacedbyeachotherinthesentence：thecomplementsindicatetherelationsbetweenthenounsandtheoneswhichareinneedofthenouns’complementtomodifythemeaningoflatter．AndinEnglishdependencyannotationSOfar,nodivisioningrammaticalrelationshasbeensearchedyet．Hence，themodifierscouldbeobminedasthegroupofdeterminer，adjectivalmodifier,prepositionalmodifier,predeterminer,nouncompoundmodifier,humPricmodifier,relativeclausemodifier,participalmodifier,appositionalmodifier,possessivemodifier,infinitivemodifier,clausalmodifier,wherethedependentsallplayasthelexicalcomplementorconstraintontheheadnouns；theparallelshereonlvcontainthedependencyrelationofconjuncfiwherenounsanditsdependentsarelexicallytreatedasthesame；thelastgroupinclude6dependencyrelations：no}味in口lobject,copula，auxiliary,negationmodifier，abbreviationandcomplementizer,where40 ThesisforMasterDegreeofZhejiangUniversityChapterFourResultsandDiscussionthenounsfunction懿thelexicalandsyntacticcomplementofitsdependents．Undoubtedly,themodifiersformthebiggestassociationofthewholedependencyrelations，accountingto89．8％，inwhichthedeterminertakesupthe37％part．Theparallelsarebuiltupbyconjunct,whichemphasizedthecoordinationbetweennouns．amountingto3．7％ofthewholecases．Amongthe6dependencyrelationsofthecomplements，nounsaccompanyingwithnominalsubjectrankthefirstposition．reachingto43．3％ofthecomplements．Theamountoffrequencyofthese6dependencyrelationscomeuptO1753．6．5％ofthewholecases．4．2DependencyRelationsofEnglishNounsastheDependentToanswerthesecondresearchquestion：‘‘ThroughwhichdependencyrelationswouldallowtheEnglishnounstobegovernedbyotherwordclasses?Whatistheproportiontheserelationsandhowtheydistribute?'’Anotherstatisticalsurveyalsohasbeenconductedinthisphase．Firstly,thepart-of-speechcolumntaggedwiththestartingletterof‘‘N”，whosecorrespondinggovernorisnotdefaultwillbeselected，andthesenounsfunctionasthedependentinthesentence．Inthenextstep，thedependencyrelationsofthesenouns硒thedependentwillbecalculatedbytheExcel．4．2．1ProbabilityofDependencyRelationsGoverningEnglishNounsAllthedependencyrelationsofEnglishnounsplayingasthedependentinthesentencehavebeensummeduplikethepreviouspart．13dependencyrelationsCanrepresentallthesyntacticfunctionsperformedbynouns嬲thedependent．Table4．3showsthesedependencyrelationsdistribution，theoccurringfrequencyofthesedependencyrelationsandtheirproportioninthewholegroup．Throughthisanalysisofthistable，thesyntacticfunctionsofnounsinthesentencecanbebetterextracted．Table4．3DependencyRelationsDistributionofNouns嬲DependentDependencyFrequencyProportionSampleSentencefromtheTreebankl沁lations(呦41 旦!!!!鱼!坚竺竺里!!竺箜21圣!受!竺曼旦翌!!!竺!篁￡!!f!!：竺曼!竺垦!!竺!垒竺!旦!!!竺!!12翌nominalsubject374524．755Theshipisgoingunderthebridge．nsubj(going，ship)directobject360024．451Shealwayseatsherlunchatnoon．dobj(eats，lunch)objectofa317520．989Andwhereareyouatthetime?prepositionpobj(at，time)nouncompound222314．695Thereisacarracenearourtowneverymodifieryear．nn(race，car)conjunct9956．577FranksmiledwhenherememberedhishardearlyyearsandthelongroadtoSuccess．conj(years，road)passivenominal5383．556Hisbicyclehadbeenfound．subjectnsubjpass(found，bicycle)temporal3262．155DebbieHartisgoingtoswimacrossthemodifierEnglishChanneltomorrow．tmod(swim，tomorrow)agent1831．011Thedeadaresaidtoreturntotheirhomesandtheyalewelcomedbytheliving．agent(weicomed，living)clausal1380．714HefeelsthathisriseinstatusiSwellcomplementworththelossofmoney．ccomp(feels，loss)indirectobject1160．635Wetelltheriveralloursecrets．iobj(tell，river)adverbialclausal460．304IaskedapassengerifthiswasthemodifierWesthavenExpress．advcl(asked，Express)42 里竺堡垒!坠竺竺里!曼坚兰21兰堕墅竖旦翌堡竺堕一．曼!呈巴!兰!竺垦!!兰坐竺璺里!!竺墅竺meaSUre3l0．205WeinfcIrmedhimthatthefissureextendedforfiftyyearsandwastwofeetwideandfourfeetdeep．measure(wide，feet)；measure(deep，feet)parataxis120．079Anothergroupwasdisturbedforanequalperiodoftimebutwhentheywerenoexhibitingeye·movements．parataxis(disturbed，eye-movements)Toml15128100TheprimarysyntacticfunctionsofEnglishnounsarelistedasfollow：subjectin24．8％，objectin24．5％，prepositionalcomplementin21％，attributivein14．7％(nouncompoundmodifier)，subjectofapassiveclausein3．6％(passivenominalsubject)，temporaladverbialin2．2％(temporalmodifieO．complementofapassiveverbin1％(agent)，predicateofaclausein1．1％(clausalcomplement,adverbialclausalmodifier,parataxis)，indirectobjectin0．6％，complementofquantifierin0．2％(measure)ofthewholecases．Theresultactuallydemonstratesthemaindistributionofnouns’syntacticfunction．ThesimilaritiesanddifferencesbetweenthisresearchandpreviousEnglishtheorieswillbeanalyzedinsubsection4．3．Inthetableabove，itcanbeobservedthattheoccurringfrequencyofnounsassubjectiscloselyassanleasthatofobject．Thesumtotalofthesetwocasesaccountsforalmosthalfofthea11cases．Also，itcanbeinferredthatnounsfrequentlyfunctionastheobjectofthepreposition，whichisalsocalledtheprepositionalcomplement．Thisfunctiontakesuponefifthofthewholesyntacticfunctionsofnouns．Thefourthlargestgroupcomesfromtheattributive，whichalsoareregardedasthenounphrasemodifier．Forexample，inthesentenceof“Thecarpenterfixedthebrokentableleg”，tablemodifiesthemeaningofleg，alsoservesfortheattributiveoftheleg．Thedependencyrelationsofpassivenominalsubjectandagentindicatethenouns’functionasthesubjectandcomplementofapassiveverb．Thepassiveoccasion43 !!!!!!尘!坚竺墼21望!121圣!笪!竺墨望21竺l旦竺塾翌墼翌坚坠!竺!竺竺壁里!!!竺!!竺happenatthe4．6％rateofthewholefunctions．Englishnounsalsocanserveastheadverbials，butonlythetemporaladverbialhasbeensortedfromthetreebankwithacomparativelyhighoccurringfrequency,directionaladverbialhappensinacomparativelylowfrequency．Besidesthedirectobject,nounsalsocallfunction嚣theindirectobjectinthesentence，forinstance，Jimmyboughtmeflowers，where“me’’actstheindirectobject,and“flowers’’船thedirectobject．TheroleofpredicateisalsoaprimaryfunctionofEnglishnouns．nethreesituationsofclausalcomplement，adverbialclausalmodifierandparataxisbothreflectthenouns’functionasthepredicate，precisely,thepredicateofclause，accountingfor1．1％ofthedistribution．Asmallproportionofnounsworkingasthecomplementofthequantifiercouldbefoundinthedependencyrelationsdistribution，similarwiththeoccaSionofnouns硒thegovernor．4．3TypicalSyntacticFunctionsofEnglishNounsThroughtheanalysisofbothsituationswherenounsfunction雒thegovernoranddependentinthedependencyrelations，thereportaboutthesyntacticfunctionofEnglishnounscanbedelivered，whichwillbecomparedwiththepreviousstudyinEnglishnounsornounphrases．Aboundaryissettodistinguishthetypicalfunctionsfromatypicalfunctions．Thefrequencyofcertaindependencytypeover100isregardedasthetypicalfunctions；others，atypicalfunctions．ThefrequencystandardissettoensurethatthecommondependencyrelationsselectedfromtreebankcancoveralltheEnglishnouns’typicalfunctionsprovidedbyEnglishgrammarofHuddleston．ARertheintegratedconsiderationofthetwotablesobtainedfromthedependencytreebank,thetypicalsyntacticfunctionsofEnglishnounsandnounphrasesinclude：subject,directobject,attributive，prepositionalcomplement，modifierofquantifier,conjunct，predicate，subjectofapassiveclause，complementofapassiveverb，adverbial，indirectobject,appositive，possessivemodifier．Amongthesefunctions，theattributive，modifierofquantifier(numericquantifier)，conjunctandpredicate(includingthepredicateofclause)couldbothbefoundindependencyrelationsof44 dependencytypesofcomplementizer,adverbialclausalmodifier,clausalcomplement，parataxis．Exceptfromthesefourcommonfunctions，thenouns’functionassubject,directobject，prepositionalmodifier(objectofpreposition)，subjectofapassiveclause，complementofapassiveverb，adverbial，indirectobjectcouldonlybefoundintherelationofnouns舔thedependent；whiletheappositiveandpossessionmodifierareexclusivefortherelationofnounsasthegovernor．Theatypicalfunctionsofnounsinclude：theheadnounofanabbreviation，theheadofrelativeclausemodifyingthenouns，andSOon．Huddleston(1984，2002)proposedtenprincipalgrammaticalfunctionsofnounsandnounphrases：subject，predicatenominative，directobject,objectcomplement，indirectobject，prepositionalcomplement,nounphrasemodifier,possessivemodifier,appositiveandadverbial．Excepttheobjectcomplement,allotherfunctionscanbefoundintheanalysisofdependency仃eebank．Thefunctionofobjectcomplementhasn'tbeendefinedbyChineseandEnglishtypes(seeAppendixIandII)，suchasthesentence：Hetreatsourpuppyourbaby,‘‘ourbaby’’playsasthecomplementoftheobject‘‘ourpuppy”．BycomparingthetypesformulatedinEnglishandChinese，somereasonsfortheabsenceoffunctionofobjectcomplementmaybepresumed．Inthemanifestationpattern，theobjectcomplementiseasilyconfoundedwithattributive(nouncompoundmodifieO，indirectobject,appositive(nouncompoundmodifieO．Thesetypesareproducedtoapplyintotheidentificationofparser．Asfortheautomaticallyoperatingsoftware，therepresentationofobjectcomplementisprobablymisunderstoodastheattributiveorindirectobjectwithoutthesemantictagsinafurtherlayer．Forexample，inthesentence“Thechilddrewhismotherapicture”，noun“hismother'’functionsastheindirectobject．and“apicture’’worksaSthedirect0biect；inanothersentence“weconsiderourdogourson”，“ourson”works舔theobjectcomplement，thestickingoftwonounphrasescouldn’tbedirectlydiscriminatedbetweenthetypeofindirectobjectandthe锣peofobjectcomplement．Anotherexampleislistedinthesentence“AmericarecentlyelectedBarackObamapresident”，45 !!!!!!121坚竺竺里!壁121圣!望!竺墨望里!：!翌堂曼塾!￡竖12竺坠!竺!堡竺12堡竺!!!竺“president’’isconsidered懿theobjectcomplement；intheothersentence“BarackObamapresidentwillvisitChinatoday”，“president'’functions邪theappositiveof“BarackObama’’．Uponthat．thedependencytypeofobjectcomplementonlycailbeaddedwiththeassistanceofsemantictagsinthefurtherstudy．Figure4．IPVPGraphofEnglishNounsasGovernor Figure4．2PVPGraphofEnglishNouns孙DependentBasedon廿lefindingsabove，thesyritacticfunctionsofEnglishnounscalculatedbythisresearchconformtothelawsmadebypreviousstudyonnounsinlargeextent·Consideringthefactthattheobjectelementhasn’tbeendefinedindependencyrepresentationrules，andactualperformanceofnounsisoutofthesummarizationmadebylinguists，morespecificrulesareexpectedtoinvolveboththedependencyrepresentationandsyntacticfunctionsofpart-of-speech·AccordingtoPVPtheory(Liu&Feng，2007)，thefrequencyofonecertainpart．of-speechisdirectlyproportionaltotheweight／intensityofthispart_of-speechwhichisalsodirectlyproportionaltotheadhesionofthepart-of-speech．Duetolackingofprecisedata,theydrawlinestoimplytheintensityofthepart。of-speech·InSteadofdrawingwithstraightlines，abarchartisusedheretoindicatetheoccurringprobabiIi锣ofonespecificdependencyrelations，thelongertheiengthofbar,themoreffequencythisdependencyrelation(syntacticfunction)isobserved·Figure47 !!!!!!堕坠箜竺旦望竺21兰塑!竺曼望翌!∑墅型篁垒!旦坚12竺坠!旦!堡竺璺里!!竺!!!竺4．1and4．2arethederivationaldependencypatternsoftheresultsillustratedinTable4．1and4．3separately．CombiningthesyntacticfunctionsdisplayedinthesituationwherenounsplayasgovernorwiththesyntacticfunctionsperformedinthesituationwherenounsplayasdependenLthemaindifferencesandsimilaritiesbetweenpresentresearchandprevioustheoreticalresearchonEnglishnounshavebeenlistedinTable4．4．Table4．4SimilaritiesandDifferencesofPreviousStudyandPresentStudyonEnglishNounsSimilaritiesDifferencesEnglishnounscanbemodifiedbyPreviousPresentStudy1．DeterminerStudy(includingpredeterminer)EnglishnounsCanbemodified2．Adjectivalmodifierby3．Attributive1．Prepositionalmodifier(nouncompoundmodifieO2．Nominalsubject4．NumericmodifierEnglish3．Copula5．Preconjunctnouns4．Infinitivemodifier6．RelativeclausemodifierCallalso5．Negationmodifierfunction6．AuxiliaryEnglishnounscanfunctionasaS7．Complementizer1．Subjectobject8．Abbreviation2．ObjectcomplemEnglishnounscanalso3．Objectofapreposition-entfunctionas4．Attributive(nounphrasemodifier)1．Subjectofapassiveclause5．Indirectobject2．Objectofapassiveverb6．Possessivemodifier3．Clausalcomplement7．Predicate4．Adverbialclausalmodifier8．Appositive5．Parataxis9．Adverbial6．Quantifier48 ThesisforMasterDegreeofZhejiangUniversityChapterFourResultsandDiscussion4．4ComparisonbetweenEnglishNounsandChineseNounsTheanswerofthelastresearchquestioncanbegivenbymeansofthecomparisonbetweentheresultsobtainedfromEnglishnounsresearchandChinesenounsresearch．TheChinesenouns’findingshereareadoptedfromtheresearchofGaoSong(2010)，whichalsoinspiresthelaunchingofthisstudy．Whennounsplayasthegovernorinthedependencyrelations．thenounscanbemodifiedbyattributer,punctuation，epithet,coordinatingadjunct,adverbial，particleadjunct,clauseadjunct,subject,pluralcomplement,adjunctofsentenceend，complementofclassifier,correlativeadjunctandparenthesisinChinesestudy；whileinEnglishstudy,thenounscanbemodifiedbydeterminer,adjectivalmodifier,prepositionalmodifier,predeterminer,nouncompoundmodifier,conjunct,numericmodifier,nominalsubject,copula,relativeclausemodifier,preconjunct,possessionmodifier,infinitivemodifier,auxiliary,negationmodifier,complentizer,clausalcomplement,abbreviation．Underthiscircumstance，someobvioussimilaritiesanddifferencesbetweenChineseandEnglishresearcheshavebeenconcludedinTable4．5．Table4．5SimilaritiesandDifferencesofChineseandEnglishNounsasGovernorSimilaritiesDifferencesNounscanbothbemodifiedbyChineseEnglish1．AttributiveChinesenounsCanEnglishnouscanbe2．Appositivebemodifiedbymodifiedby3．Conjunct1．Adverbial1．Otherdeterminers4．Clausalcomplement2．Particleadjunctexceptforpost-determiner5．Subject3．Adjunctof2．Adjectivalmodifier(nounshereworkaspredicate)sentenceend3．Prepositionalmodifier6．Post-determiner4．Parenthesis4．Copula7．Numericqualifier5．Relativeclause8．Coordinationmodifier6．Preconjunct7．Participalmodifier8．Appositionalmodifier49 ThesisforMasterDegreeofZhejiangUniversityChapterFourResultsandDiscussion9．Possessionmodifier10．Infinitivemodifier11．Auxiliary12．Negationmodifierl3．AbbreviationinChinesenounsthattheadverbialscanbedirectlylinkedwithnounstostandfortheadjective，andthisphenomenoncan’tbesearchedinEnglish，forinstance，inthesentencen“他在电视上都讨论很学术的问题”，theadverbial“很”directlymodifyingthenoun“学术”tosignifytheadjective“academic”．Theparticleadjunctstandsforaparticle“等”followinganoun，whichequalstothe“etc”inEnglish，likethesentence“他们参加了篮球、足球等运动”．TheadjunctofsentenceendalwayssignifiestheexclamatoryorinterrogativesentenceinChinese，andthiscouldn’texistinEnglish，forEnglishwoulduseinterrogativewords，likewh-groupandhow,toimplytheexclamatoryorinterrogativesentence．Theseadjunctsofsentenceendusuallyarethemoodparticles，like”呢”，”啊”，”着呢”，”吧”．AndChinesenounsalsoCanbemodifiedbytheparenthesis，whichhasn’tbeendefinedbyEnglishdependencyrepresentation，inthesamplesentence“比方说，你的苦恼”，“苦恼”ismodifiedbytheparenthesis“比方说”．TheprimarysyntacticfunctionsofEnglishnounshaven’tbeensortedfromChinesenouns’functionare：theEnglishnounscanaccompanywithprepositionalmodifier,predeterminer,preconjunct,possessionmodifier,infinitivemodifierandabbreviation．ApartfromthesituationthatEnglishandChinesenounsconnectingwithsubjectactasthepredicateinthesentence，Englishnounsthatgovernorthecopulaverborauxiliarywillalsoperformthepredicateinthesentence．ThenounsfromthesesthreetypesCallallbemodifiedbythenegationdirectly．Whennounsplayasthedependentinthedependencyrelations，commonground ThesisforMasterDegreeofZhejiangUniversityChapterFourResultsandDiscussiononnouns’syntacticfunctionhasbeenachieved．InChinesestudy,nounscanbepresentedastheobject，subject，attributer,prepositionalobject，complementof‘de’，postpositionalcomplement,adverbial，epithet，subobject,topic，maingovernor，coordinatingadjunct,parenthesis，indirectobject,sententialobject；whileinEnglishstudy,nounscanbepresentedasnominalsubject，directobject,objectofapreposition，nouncompoundmodifier,conjunct，passivenominalsubject,temporalmodifier,agent，clausalcomplement，indirectobjecLadverbialclausalmodifier,measure，parataxis．Underthiscircumstance，someobvioussimilaritiesanddifferencesbetweenChineseandEnglishresearcheshavebeenconcludedinT{出le4．6．Table4．6SimilaritiesandDifferencesofChineseandEnglishNounsasDependentSimilaritiesDifferencesNounscanbepresentedasChineseEnglish1．SubjectChinesenounsalsocanbeEnglishnounsalso2．Objectpresentedascanbepresentedas3．Attributive1．Complementof‘de’1．Subjectofa4．Objectofapreposition2．Postpositionalpassiveclause5．Temporaladverbialcomplement2．Objectofapassive6．Appositive3．Subobjectverb7．Conjunct4．Topic3．Clausal8．Indirectobject5．Parenthesiscomplement9．Predicate6．Sententialobject4．Adverbialclausalmodifier5．Parataxis6．QuantifierDiffcrcnceslieinChinesenounsincludethesyntacticfunctionofcomplementof‘de’，whichactuallyequaltothedependencytypeofpossessionmodifierplayedbyEnglishnouns，reflectthegenitiveofnouns，forinstance，inthesentence“人民的利益是最需要考虑的”，“人民”worksasthecomplementof”的”．PostpositionalcomplementinChineserepresentsthatChinesecanbefollowedbyapostpositiontoemphasizethescopeofthenoun，inthesentence“国际上取消了对食用味精限制的5l !!竺!!竺坠竺坚望!受121兰!复!竺墨望里!：翌堕g!!里坚12竺垦!!竺!竖竺璺2堡!竺!12翌规定”，though“国际”isfollowedbyapostposition“上”，itdoesn’tinfluencethenountoworkasthesubject．Besides，ChinesedependencytypesraisedthespecificdefinitionforTopic“事”。whichisequivalenttoapropernounforEnglish．ThedependencytypesofSubobjectandSententialObjecthaven’tbeenmentionedinEnglishindependently．Thedistinctionbetweensubobjectandsententialobjectliesinwhetherthesubordinationclauseissetinoroutofthemainclause．such舔thesubobject“人民”inthesentence“我们要带领人民从新的历史起点出发”，andthesententialobject“年纪”inthesentenee“您看您这么大年纪了，还继续工作呢”．ThemaingovemorofChinesestudyalsoplaysasthepredicateinEnglishstudy,whichcanbesearchedoutasaform“predicateofclause”ofclausalcomplement，adverbialclausalmodifierandparataxisinEnglishdependencyrelations．Particularly,Englishnounscanworkasthepassivenominalsubjectandagent，theformerofwhichindicatesthenominalsubjectofpassiveclause，whilethelattermeansthecomplementofapassiveverb．AndthesetwofunctionsofEnglishnounscanbelocatedintheconstructionof“bei”inChinesestudy．Forinstance，Englishsentence“Thebicycleshavebeenfound”canbetranslatedinChinese“自行车被找到了”，“bicycles”isthepassivenominalsubject；alsointheEnglishsentence‘‘Thecriminalsarearrestedbythepolice”canbeinterpretedtoChinesesentence“罪犯被警察抓住了”，thenoun‘'police”worksastheagenthere．TheprobabilitydistributionofChinesesyntacticfunctionsisalittledifferentfromthatofEnglish．ForthestatisticalresultmadeinChinesestudy,thetop5syntacticfunctionsarerankedasthesequenceofobject,subject,attributive，adverbialandpredicate，anotherprimaryfunctionofcomplementhasn’tbeensortedintheChinesedependencytreebank；thesequenceofprimarysyntacticfunctionsofEnglishnounscalculatedinthisEnglishdependencytreebankisfollowed：subject，object,objectofthepreposition，attributive(nounphrase)，predicate，adverbial，appositive，possessivemodifierandindirectobject．52 ChapterFiveConclusion5．1ConclusionsFourresearchquestionsaleinvestigatedinthisresearch，andfourconclusionscanbedrawnfromthedataofdependencyTreebank．WithrespecttotheEnglishnounsasthegovernorofotherpart-of-speech，20dependencyrelationshavebeencalculatedfromthedependencytreebankafterautomaticannotationofStanfordDependencyParserandthemanualverificationofresearchers．Englishnounsalemostprobablymodifiedbythedeterminer,accountingforonethirdfrequencyofthewholedependencyrelations．Theassociationofdeterminer(41．4％，includingpredetermineO，adjectivalmodifier(17．5％)，prepositionalmodifier(12％)，nouncompoundmodifier(8．3％)，conjunct0．7％)，numericmodifier(2．9％)，relativeclausemodifier(2．7％)，preconjunct(1．2％)，participalmodifier(1．1％)，appositionalmodifier(1％)，possessionmodifier(1％)，infinitivemodifier(0．6％)formstheprimarymodifieroftheheadnoun．Besides，thenounsCallactasthepredicatewhentheygovernthenominalsubjectandcopulaverb．Thosenounsmodifiedbythenegationmodifier,auxiliaryandcomplementizerarealsoregardedasthepredicateofthesentence．Thenounsfunctioningaspredicateinthiscaseamountto6．4％．Therealeonly5casesin0．01％indicatingnounsbeingspecifiedbytheabbreviation．Inthecircumstanceofnounsasthedependent,13dependencyrelationsaresorted，whereEnglishnounsfunctionasthesubject(24．8％)，directobject(24．5％)，objectofapreposition(21％)，attributive(14．7％)，subjectofthepassiveclause(3．6％)，temporaladverbial(2．2％)，complementofthepassiveverb(1％)，indirectobject(0．6％)．Otheroverlappingfunctionsliketheoccasionsofnounsasgovernorcarlalsobeenobtainedhere，nouns’functionastheconjunct(6．6％)，predicate(1．1％)andquantifier(0．2％)．Throughthesummarizationofthetwopreviousstudies，thetypicalsyntacticfunctionsofEnglishnounscanbeconcludedassubject,object,objectofthepreposition，conjunct,attributive(nounphrase)，predicate，adverbial，appositive，53 ．T．．h．e．．s．i．s．．f．o．r．．M．．．a．s．t．e．r．．D．．e．g．．r．e．e．．o．．f．Z．．h．．e．j．i．a．n．．g．．U．．n．．iv．．e．r．s．i．t．y．．．．．．．．．．．．．．．．．．—．．—————————————ChapterFiveConclusion．．possessivemodifierandindirectobject．IncontrastwiththetheoreticalstudyongrammaticalfunctionsofEnglishnouns，thefindingsofthisresearchmeetthelawsconcludedbypreviousstudyonthewhole．However，thefunctionofobjectcomple：mentcouldn，tbefoundforthisrepresentationhasn’tbeenexplicitlydescribedyetSO饥觚dtheambiguityliesin’theliteralrepresentationcanonlybesolvedwiththeassistanceofsemanticannotation．ThelaststepondataanalysisisimplementedinthecomparisonofChinesenounsresearchandEnglishnounsresearch．Generally,theysharemorecommonnessthandiscrepancy．Theyallusuallyfunctionassubject,object,attributive，objectofapreposition，adverbial，predicate，appositive，conjunct，indirectobject,beingmodifiedbvtheclausalcomplement,numericmodifiers，coordination，andtheyCanallreflectthepassivevoice．Butthedifferencealsocanbelocated，liketheChinesenounscanbemodifiedbyadverbialdirectlytospecifytheadjective；thedependencytypesofadjunctofsentenceend，parenthesis，topic，postpositioncomplementareexclusiveforChinesenounswhichcomplywithcharacteristicofChinesenouns．ThefunctionofEnglishnounsbeingmodifiedbyprepositionalmodifier,predeterminer,preconjunct,andabbreviationaren’tcontainedbyChinesedependencytypes．ThevarioustypesfordeterminersofEnglishnounsornounphrasescouldonlyrespondtothepluralcomplementofChinesenoun．5．2LimitationsandSuggestionsGroundedonthePVPtheory,thisresearchaimedtoveerythesyntacticfunctionsofEnglishnounsintheassistanceofEnglishdependencytreebank．Thedataareallextractedfromtruematerialswithhi【ghreliability．Comparedwiththerationalinvestigationonpart．of-speech，theactualperformancerealcontextwillauthenticallyreflectthefrequentusageofpart-of-speechinthesentence．AssumingthattheresultofthisresearchwouldbeputintopracticeduringtheEnglishteachingprocess，theieameI'Scouldbettercomprehendthesyntacticfunctionsintheorderoftheirimportanceoroccurrence．Despiteofgreatexpectation，therestillexiasomeinadequateaspectsingreatneed54 ’J[hesisforMasterDegreeofZhejiangUniversityChapterFiveConclusionofenhancement．Thebiggestconstraintcomesfromtheadoptionofunifieddependencytypes，thoughthesyntacticrepresentationsforPennTreebankandStanfordDependencyParserhasbeenappliedinthisresearch，thevacancyofsomedependencyrelations，suchastheobjectcomplement,stilldemandsacomprehensivefilling．Moreparticularcasesbeingconsideredanddetailedspecificationofdependencyrelationsbeingprovided，moreaccuratesyntacticstructurescanbepresented．Theincorrectmanualannotationformsthesecondbiggestlimitationinthisresearch，fortheverificationtaskisallcompletedbytheresearcher,theaccuracyofthisannotationiscomparativelylowthanthemanualannotationdonebymoreresearchers．111enegotiationandcooperationbetweentheresearchersshouldbeinvolvedinthefuturestudy．Anotherrestraintofthispresentresearchisfromtherawmaterialsandsizeofcorpus．Currentmaterialsaremostlyconcemingwithexpositivecomposition，andithasresultedinthedefaultofsomesyntacticfunctions，forinstance，nounsasthepostpositionadverbialhaven’tbeenretrievedfromthistreebank．Thus．thematerialsinmoregenresshouldbeaddedintothecorpus，alsothesizeofcorpusneedstobeenlargedforthestudyonotherparts-of-speech．55 ThesisforMasterDegreeofZhoiangUniversityRefcrelip．．宅sReferencesAbeill6，A(cd)．Treebanks：BuildingandUsingParsedCorporaIM]．KluwerAcademicPublishers，2003．Agel，Vilmos．Valenzrealisierung，GrammatikundValenz：GrammarofValency．In：ZGL，1995，23：2-32Allenon，D．J．ValencyandtheEnglishVerb[M】．London；NewYork：AcademicPress，1982．Bresnan，J．Lexical-FunctionalSyntax【M】．Oxford：BlackwellPublishers，2001．By,Tomas．EnglishDependencyGrammar．InAdvancesinDependencyGrammar,WorkshopCOLING，2004．Carroll，J&Minnen，GandBriscoe，T．CorpusAnnotationforParserEvaluation．InProceedingsoftheEACLWorkshoponLinguisticallyInterpretedCorpora,1999．Charniak,E．AMaximum·Entropy—Inspired—Parser．InProceedingsofNAACL，2000．Charniak,E．Immediate—HeadParsingforLanguageModels．InProceedingsofACL，2001．Chomsky,N．LecturesonGovernmentandBinding【M】．WakerDeGruyterInc，1993．Collins．M．Head—DrivenStatisticalModelsforNaturalLanguageParsing．Ph．D．Thesis，UniversityofPennsylvania,1999．Corazza,A．&Satta．G．ProbabilisticContext-FreeGrammarsEstimatedfromInfiniteDistributions【刀．IEEETransactionsonPatternAnalysisandMachineIntelligence，2007：1379—1393．Covington，M．A．AFundamentalAlgorithmforDependencyParsing．Proceedings，39mAnnualACMSoutheastConference，Athens，Georgia,2001．Cresswell，R．&Szekely,C．eta1．ADependencyTreebankforEnglish．ThirdInternationalConferenceonLanguageResourcesandEvaluation，2002：857—863． ，‘一⋯__⋯⋯_“ThesisforMasterDegreeofZhejiangUniversityReferencesDebusmann，R．&Kruij危G．J．ExtensibleDependencyGrammar：ANewMethodology．InProceedingsoftheWorkshoponRecentAdvancesinDependencyGrammar,2004：7&_85．Eisner,J．M．AnEmpiricalComparisonofProbabilityModelsforDependencyGrammar．TechnicalReport．UniversityofPennsylvania,1996．Emons，R．ValenzgrammatikFiarDasEnglische．Ttibingen：ValencyGrammarofEnglish[M】．Niemeyer,1978Engel，U．DeutscheGrammatik：GermanyGrammar[M]．北京语言学院出版社，1992．Frankland，E．OnaNewSeriesofOrganicBodiesContainingMetals【M】．PhilosophicalTransactionsoftheRoyalSociety,1852．142：417州．FunctionalGenerativeDescription，fromWikipedia,thefreeencyclopedia,Http：／／en．wikipedia．org／wiki／Functional-Generative_DescriptionGaifman，H．DependencySystemsandPhrase-StructureSystems阴．InformationandContr01．1965．8：304-307．Giorgi，A．neSyntaxofNounPhrases【M】．England：CambridgeUniversityPress，1991．Hays，D．G．DependencyTheory：AFormalismandSomeObservations【J】．Language，1964．40：511-525．HeUwig，PDependencyUnificationGrammar．Proceedings，COLING，1986：195—198．Herbst，T．EnglishValencyStructures—AFirstSketch．EESE，1999．Herbst,T．&Heath，D．eta1．AValencyDictionaryofEnglish：ACorpus-BasedAnaysisoftheComplementationPatternsofEnglishVerbs，NounsandAdjectives．Berlin／NewYork：MoutondeGruyter,2004．Huddleston，R．IntroductiontotheGrammarofEnglish眦】．Cambridge：CambridgeUniversityPress，1984．57 RefeI'ellCe$Huddleston，R．&Pullum．eta1．TheCambridgeGrammaroftheEnglishGrammar[M】．CambridgeUniversityPress，2002．Hudson，R．A．EnglishWordGrammarIM]．Oxford：Blackwell，1991．Hudson，R．A．LanguageNetworks：TheNewWordGrammar【M】．OxfordUniversityPress，2007．Jlirvinen，T．&Tapanaien，P．TowardsanImplementableDependencyGrammar．InProceedingsofCOLING-ACL．Workshop：ProcessingofDependencyGrammar,1998．Jurafsky，D．&Martin，J．SpeechandLanguageProcessing-AnIntroductiontoNaturalLanguageProcessing：自然语言处理综论．(冯志伟．孙乐译)．北京：电子工业出版社。2005．Karlsson，E&Voutilainen，A．(eds．)ConstraintGrammar：ALanguage·IndependentSystemforParsingUnrestrictedText．Berlin：MoutondeGruyter,1995．Kruijff．G．J．M．FormalandComputationalAspectsofDependencyGrammar：HistoryandDevelopmentofDG，TechnicalReport，ESSLLI，2002．Kruijif,G．J．M．DependencyGrammar．InKeithBrown(ed．)TheEncyclopediaofLanguageandLinguistics．ElsevierPublishers，2006．Lai，B．&Huang，C．N．AnApproachtoDependencyGrammarforChinese．StudiesinChineseLinguistics．HongKong：LinguisticSocietyofHongKong，1998：143-163．Lin，D．OntheStructuralComplexityofNaturalLanguageSentences．Proceedingsofthe16血InternationalConferenceonComputationalLinguistics(COLING)．1996：723．728．Lin，D．Dependency-BasedEvaluationofMINIPAR．InWorkshop011theEvaluationofParsingSystems，Granada,Spain，1998．Liu，H．T．&Huang，W：AChineseDependencySyntaxforTreebanking．IntheProceedingsofthe20mPacificAsiaConferenceonLanguage，InformationandComputation，Beijing：TsinghuaUniversityPress，2006：126-133．53 一ThesisforMasterDe—g—r—e—e——o—f—Z——h—e—j—ia—n—g——U——n—iv—e—r—s—i—ty——．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．．R．．e—f—e—re．n．．ce—．s．一一—————————————————----—--------__●___●___●__________●-__●_--___-__-__●__●_●_●_●__●____-________●____●_●__--．．．-．．．．——一————Liu，H·T．DependencyDirection船aMeansofWord．OrderTypology：AMethodBasedonDependencyTreebanks【J】．Lingua,2010，120：1567．1578．Manning，C·D．&SchUtze．H．FoundationsofStatisticalNaturalLanguageProcessing【M】．TheMITPress，1999：199．226．Mameffe，M．C．&MacCartney,B．GeneratingTypedDependencyParsesfromPhraseStructureParses．InLREC，2006．Mameffe，M．C．&Manning，C．D．StanfordTypedDependenciesManual．StaIlf．0rdUniversity,2008．Maruyama,H．StructuralDisambiguationwithConstraintPropagation．IntheProceedingsoftheAnnualMeetingofAssociationforComputationalLinguistics，1990：3l-38．Meaning。TextTheory,fromWikipedia,thefreeencyclopedia,Http：／／en．wikipedia．org／wiki／Meaning—TextTheoryMel’6ul(，I·AFormalLexiconintheMeaning—TextTheory．ComputationalLinguistics．1987：261．275．Menzel，W：&SchrSder,I．DecisionProceduresforDependencyParsingUsingGradedConstraints．ProceedingsoftheWorkshoponProcessingofDependency．BasedGrammars，1998：78--87．McDonald-R·&Pereira,EOnlineLearningofApproximateDependencyParsingAlgorithms．InProceedingsofEACL，2006：81．88．Nivre，J·DependencyGrammarandDependencyParsing．Citeseer,2005Nivre，J．InductiveDependencyParsingofNaturalLanguageText．Ph．DThesis，Vaxj8University,2005．Prasad，R&Dinesh，N．eta1．ThePennDiscourseTreeBank2．0．UniversityofPennsylvania，2007．ProbabilisticContext—FreeGrammar,fromWikipedia，thefreeencyclopedia,59 !璺竺垫鱼!坚箜堡!旦!￡堡121兰皇笪!竺墨望里!：!翌!堡坠垒堡里竺!Http：／／en．wikipedia．org／wiki／Probabilisticcontext—free_grammarQuirk,R&Sidney,G．eta1．AComprehensiveGrammaroftheEnglishLanguage[M】．LondonandNewYork：Longman，1985：241-332．Sgall，P．Dependency—BasedFormalDescriptionofLanguage阴．IntheEncyclopediaofLanguageandLinguistics．Oxford：PergamonPress，1994，2：867-872．Sleator,D．D&Temperley,D．ParsingEnglishwithaLinkGrammar．InThirdInternationalWorkshoponParsingTechnologies，1993．Tallerman，M．PhraseStructureVS．Dependency：TheAnalysisofWelshSyntacticSoftMutation【J】．Linguistics．CambridgeUniversityPress，2009：167·201．Tesni6re，L．1116mentsdeSyntaxeStructurale[M】．Paris：EditionsKlincksieck,1966．TheStanfordParser．TheStanfordNaturalLanguageProcessingGroup，2006．Http：／／nlp．stanford．edu／software／lex-parser．shtmlWang，W-&Harper,M．PAStatisticalConstraintDependencyGrammar(CDG)Parser．In42“dAnnualMeetingoftheAssociationforComputationalLinguigics(ACL)WorkshoponIncrementalParsing，2004．YangHuizhong，OpeningAddressonthe2003InternationalConferenceonCorpusLinguistics，Shanghai，2003．陈波．基于依存语法的语料库标注研究．第七届中文信息处理国际会议．2007：286．289．北京语言学院语言教学研究所．汉语词汇的统计与分析【M】．外语教学与研究出版社．1985．冯志伟．现代语言学流派[M】．陕西人民出版社．1989：175．184．冯志伟．自然语言的计算机处理【M】．上海外语教育出版社．1995：1．13．冯志伟．基于经验主义的语料库研究[M】．术语标准化与信息技术．2007．冯志伟．从属关系语法的某些形式特性．教育部语言文字应用研究所．2009． 1hesistorMasterDegreeotLhejiangUnwersityReterences高松，赵怿怡，刘海涛．汉语特殊结构的句法标注策略，中国计算语言学研究前沿进展(2007-2009)，清华大学出版社．2009高松．基于依存树库的现代汉语名词语法功能的计量研究叽．华文教学与研究．2010，2：54-60高松．基于树库的现代汉语动词句法功能的计量研究叨．汉语学习．2010，5：105．112刘海涛．依存语法和机器翻译【J]．语言文字应用．1997，3：89-93刘海涛．自然语言处理的概率配价模式理论【J】．语言科学．2007，3：32-41刘海涛，赵怿怡．基于树库的汉语依存句法分析阴．模式识别与人工智能．2009，22：17—21．马金山．基于统计方法的汉语依存句法分析研究．工学博士学位论文．哈尔滨工业大学，2007．马永军．基于依存语法的自然语言处理现状及前景展望[J】．学术交流．2007，10：137-140莫彭龄，单青．三大类实词句法功能的统计分析【J】．南京师范大学学报．1985，3．沈阳，郑定欧．现代汉语配价语法研究【M】．北京大学出版社．1995．袁毓林，郭锐．现代汉语配价语法研究2【M】．北京：北京大学出版社．1998．詹卫东．一个汉语语义知识表达框架：广义配价模式．1999赵怿恰，高松，刘海涛．基于依存语法的汉语并列结构自动分析研究，中国计算语言学研究前沿进展(2007-2009)．北京：清华大学出版社，2009：148-153．赵怿怡，关润驰．汉语依存树库的构建，第三届学生计算语言学研讨会论文集．2006：438．441．郅友昌，刘万义．配价语法与配价语义补偿[J】，解放军外国语学院院报．2000．61 卫!!堡塑坚竺堡!望!g!竺21兰垒笪!竺兰旦里!：!坚!堡————————j至!璺兰竺!!竺一周国光．汉语配价语法论略阴．南京师大学报(社会科学版)．1994．周明，黄昌宁．面向语料库标注的汉语依存体系的探讨[J】．中文信息学报．1994周强．汉语句法树库标注体系【J】．中文信息学报．2004，4．朱德熙．语法答问[M】．商务印书馆．1985． AppendixIListofStanfordTypedDependencyRelationsAbbreviationImplicationExampleSentenceabbreviation’’TheAustralianBroadcastingCorporation(ABC)”abbrevmodifierabbrev(Corporation，ABC)adjectivalaeomp”Shelooksverybeautiful”acomp(100ks，beautiful)complementadverbialclause”Ifyouknowwhodidit,youshouldtelltheteacher”advclmodifieradvcl(tell，know)”Geneticallymodifiedfood”advmod(modified，advmodadverbialmodifiergenetically)”Themanhasbeenkilledbythepolice”agent(killed，agentpolice)amodadjectivalmodifier’’Sameatsredmeat”amod(meat，red)appositionalappos”Sam，mybrother'’appos(Sam，brother)modifierattrattributive”Whatisthat?”attr(is，What)aUXauxiliary”Heshouldleave”aux(1eave，should)”Kennedyhasbeenkilled”auxpass(kilied，been)auxpasspassiveauxiliaryaux(killed，has)CCcoordination’’Billisbigandhonest”cc(big，and)clausal”Iadmirethefactthatyouarehonest”ccomp(fact，ccompcomplementhonest)complmcomplementizer”Hesaysthatyouliketoswim”complm(1ike，that)conjconjunct”Billisbigandhonest”toni(big，honest)copcopula”Billisanhonestman”cop(man，is)csubjclausalsubject”Whatshesaidmakessense”csubj(makes，said)csubjpaclausalpassive”Thatsheliedwassuspectedbyeveryone”SSsubjectcsubjpass(suspected，lied)detdeterminer”Whichbookdoyouprefer?”det(book,which)dobjdirectobject”Shegavemeahug”dobj(gave，hug) ThesisforMasterDegreeofZhejiangUniversityAppendixexplexpletive”Thereisacatintheroom”expl(is，There)infmodinfinitivalmodifier”PointstOestablishare⋯”infmod(points，establish)iobjindirectobject”Shegavemeahug”iobj(gave，me)”Forcesengagedinfightingafterinsurgentsattacked”markmarkermark(attacked，after)measure-phrasemeaSUre”Themanis65yearsold”measure(old，years)modifiernegnegationmodifier”Billisnotascientist”neg(scientist，not)nouncompoundnn”Oilpricefutures”nn(futures，oil)；nn(futures，price)modifiernsubjnominalsubject”ClintondefeatedDole”nsubj(defeated，Clinton)nsubjpapassivenominal”DolewasdefeatedbyClinton”nsubjpass(defeated，SSsubjectDole)nUmnumericmodifier”Sameatsthreesheep”num(sheep，three)elementofnumber”Ilost$3．2billion”number($，billion)compoundnumberparatax!parataxis”Theguy,Johnsaid，leftearlyinthemorning”Sparataxis(1eft，said)’’BilltriedtOshootdemonstratinghisincompetence”partmodparticipialmodifierpartmod(shoot，demonstrating)prepositional”Theyheardaboutyoumissingclasses”pcomp(about,pcompmissing)complementobjectofapobj”Isatonthechair”pobj(on，chair)prepositionpossession”Bill’Sclothes”poss(clothes，Bill)possmodifierpossesslpossesswe”John’Spen”possessive(John，’S)Vemodifier’’Boththeboysandthegirlsarehere”preconj(boys，preconjpreconjunctbothlpredetpredeterminer”Alltheboysarehere”predet(boys，a11)prep／preprepositional”Isawacatinahat”prep(cat，in)pcmodifierphrasalverbprt”TIleyshutdownthestation”prt(shut，down)particlepunctpunctuation”Gohome!”punct(Go，!)purposeclause”HetalkedtOhiminordertosecuretheaccount”purpclmodifierpurpcl(talked，secure) ThesisforMasterDegreeofZhejiangUniversityAppendixquantmquantifierphrase”About200peoplecametotheparty”quantmod(200，odmodifierAbout)relativeclausercmod”Isawthemanyoulove”rcmod(man，love)modifierrefreferent”Isawthebookwhichyoubrought”ref(book,which)rclrelative”Isawthemanwhoselifeyoulove”rel(10ve，wife)tmodtemporalmodifier”LastnighLIswaminthepool”tmod(swam，night)openclausalxcomp”HesaysthatyouliketOswim”xcomp(1ike，swim)complementxsubjcontrollingsubject”Tomlikestoeatfish”xsubj(eat，Tom) AppendixlIListofChineseDependencyTypesTypeLabelTypeLabelMaingovernorSSententialobjectSentObjSubjectSUBJAuxiliaryverbObjAObjectoBJCoordinatingmarkC．IndirectObjectOBJ2AdverbialAVDASubobjectSUBOBJVerbadjunctVASubjectComplementSoCAttributerA啊己PrepositionalObjectPoBJT0picToPPostpostionalComplementFCCoordinatingadjunctCOORComplementCOMPEpithetEPAComplementofusde‘的’DECNumeraladjunctMAComplementofusde‘地’DICAspectadjunct尉Complementofusde‘得’DFCAdjunctofsentenceendESAObjectofPba‘把’BaOBJParenthesisInAPluralcomplementPLCClauseadjunctCROrdinalcomplementOCCorrelativeadjunctCsRComplementofclassifierQCParticleadjunctAuxRConstructionofPbei‘被’13}eiSPunetuationPunct AppendixIIIAlphabeticalListofPart-of-SpeechTagsUsedinTreebankNumberTagDescriptionNumberTagDescriptionlCCCoordinating19PRPSPossessivepronounconjunction2CDCardinalnumber20RBAdverb3DTDeterminer21RBRAdverb，comparative4EXExistentialthere22RBSAdverb，superlative5FWForeignword23RPParticle6INPrepositionor24SYMSymbolsubordinatingconjunction7】】Adjective25TOto8J瓜Adjective，26UHInterjectioncomparative9JJSAdjective，27VBVerb，baseformsuperlative10LSListitemmarker28VBDVerb，baseformllMDModal29VBGverb．gerundorpresentparticiple12NNNoun，singularor30VBNVerb，maSSpastparticiple13NNSNoun，plural3lVBPVerb．non．3mpersonsingularpresent14NNPPropernoun，32VBZVerb，3rdpersonsingularpresent15NNPSPropernoun，33WDTWh．determinerplural16PDTPredeterminer34WPWh—pronoun17POSPossessiveending35WPSPossessivewh-pronoun18PRPPersonalPronoun36WRBWh．adverb67 Publications【l】袁茎瑜．颜色词的语义辨析，文艺生活·文艺理论，2010，10：127-128 AcknowledgementsAsoldsayinggoes，allgoodthingsmustcometoanend．Duringthethree。yearacademiccareerinZhejiangUniversity,countlesssweetmemoriestakerootinmymind，andcountlessgenuinethanksshouldbeexpressed．Firstofall，1wanttogivethesinceregratitudetomyacademicsupervisorProf．QuYunhuaforherprofessionalinstruction，rigorousscholarship，broadtoleranceandpainstakingcareonmeforthesethreeyears．Theearnestteachingsandscrupulousworkmannerinspiremetobecomearesponsiblepersonformyworkandlife，and1willbenefitfromitfortherestofmylife．SpecialthankswillgotoProf．FengZhiweiandProf．LiuHitao·Asthedistinguishedscholars，theyallgenerouslydelivertremendousguidanceandsupporttomeinprocessofthisthesispreparationandwriting．Thelastthanksshouldbeextendedtomyparentsandmyboyfriend，fortheircontinuoussupportandassistance，loveineachdayofmylife．
此文档下载收益归作者所有
