基于主题划分的微博检索-研究

《基于主题划分的微博检索-研究》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、...华中科技大学硕士学位论文AbstractThepopularonlinesocialnetworkingservices,especiallymicrobloggingservice,arechangingthewaywhichinformationdisseminatingandsharing.Microbloggingservicesallowuserspublishandreceivemessagesfreelyandinstantly.Twitterhasmorethan100millionuserspublishing230milliontweetsonasi

2、ngleday.Insteadofsubmittingtheirqueriestotraditionalsearchengines,Twitterusersaremorewillingtoretrievenews,URLsandusercommentsthroughtwittersearch.Themicroblogsearchhasbecomeanimportantchannelforuserswhowanttogethoteventinformation.Inthispaperanewmicroblogsearchplanhasbeenproposedbasedonthe

3、factthatuserswhosearchonmicroblogsaremorecaringaboutthehottopicsandevents,andthisplancanwellmeetthetopicsearchrequirements.Firstly,classifythetweetsaccordingtotheirtopics,andthencreatetopic-microblog2-levelindexbasedontheclassificationresult.Inthispaper,anewmicroblogclassificationmethodhasb

4、eenproposedbasedontheTagswithinmicroblogs.Firstly,computetheco-occurrenceinformationinthedatasetandthengetthesimilaritymatrixbyusingthemutualinformationtool.Secondly,usetheclusteringmethodstogetthetagclusters,andallocatethetweetswhichhavetagstotheaccordingclusters.Thirdly,usetheabovetweetcl

5、ustersastrainingdatatobuildatweetclassifieranddotheclassificationjobfortheremaininguntaggedtweets.Atlastalltweetswillbeallocateintheiraccordingtopics(eachclassstandsforatopic).Buildatopic-tweet2-levelindexbasedontheaboveclassificationresult.Createthetopiclevelindexbyindexingthecentroidvecto

6、rforeachtopic,andbuildthetweetindexwithineachtopic.Inthesearchingphase,submittheuserquerytothetopicindextofindtherelevanttopicsanddothescoringforeachretrievedtopic.Afterthat,searcheachrelevanttopictofindtherelevanttweets.Thefinalscoreforeachtweetisthescoreofthetopicwhichthetweetbelongstomul

7、tiplythestatictweetscoreinthetopic.Atlast,ranktherelevanttweetsaccordingtotheirfinalscores,andreturntherankedtweetstotheuser.TheexperimentsandtestsontheTwitterdatasethaveshownthefeasibilityandeffectivenessofthisplan.Accordingtothegeneralevaluationmethods