资源描述:
《基于维基百科的搜索引擎检索结果聚类》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、上海交通大学硕士学位论文基于维基百科的搜索引擎检索结果聚类姓名:史天艺申请学位级别:硕士专业:计算机应用技术指导教师:李明禄20090101上海交通大学硕士学位论文摘要突出阐述了特征项的抽取方法和基类合并规则。在特征项抽取过程中,在短语层次上采用全二分最大长度匹配快速识别主要词汇,对同义词进行归并,对多义词根据上下文语境和语义信息的线性融合进行歧义消解。采用倒排索引的方式组织基簇,在基簇的基础上,根据簇文档之间的重叠程度、簇标签之间的上下位关系和语义相关度,采用自底向上的层次聚类算法将标签组织为层次型标签树。最后,实现了在线聚类系统Wiki-SRC,并将它与其他相关研究成果进行了比较。
2、通过对比实验,我们发现Wiki-SRC在簇标签质量、簇覆盖率等方面都有所改进,从而证明了该算法的可行性和有效性。关键词:搜索引擎,检索结果,聚类,维基百科,知识库第II页上海交通大学硕士学位论文ABSTRACTClusteringofSearchResultsBasedonWikipediaKnowledgeABSTRACTSearchEngineprovidesawayforpeopletomakefulluseofrichresourcesonInternet,however,thehugeamountofwebpagesandquerynoteasytoexpressmaking
3、toomanyreturnedsearchresultsatatime,soeffectiveorganizationofthesesearchresultsiscriticalforanysearchengine.Nowadays,mostsearchenginewillreturnalongflatlistofrecordsthataresortedaccordingtotheircorrelationwiththequery,whichrendersthemanualsearchofactuallydesiredinformationdifficult,especiallyfor
4、ambiguousqueries.Ifsearchresultscanbegroupedonthebasisoftheircorrespondingsemanticinformation,itwillfacilitateausertobrowseintoresultsofinterestquickly.However,mosttraditionalsearchresultclusteringalgorithmreliesonwordfrequencyandstatisticsinsteadofdocumentinherentsemanticinformation,oftengenera
5、tesinaccurateoutput.Inthisthesis,wepromoteasearchresultclusteringalgorithmbasedonWikipediaKnowledge.Thisalgorithmutilizesextractedsemanticrelationsbetweenphrasestoformmorecompactcluster,moredescriptiveclusterlabelandmorereasonablehierarchicalclustertree.Thisthesisfirstintroducesanovelmethodtomin
6、eknowledgefromWikipedia.ByanalyzingtheframeworkofWikipediawebsite,especiallyitshyperlinks,combinedwithsomemachinelearningmethods,heuristicrulesandstatistics,wegetathesauruswithfullcoverageandhighaccuracyaswellassomeimportantindicatorslikephrasesemanticsimilarity,phraseimportanceetc.Integratingth
7、eextractedknowledgeintokeyphraseextractionandtextclassificationshowsbetterperformance.第III页上海交通大学硕士学位论文ABSTRACTAfterwards,themainideaofsearchresultclusteringbasedonWikipediaKnowledgeisdescribedindetail,highlightingthefeature