基于主题模型和混合模型的微博客交叉话题发现的研究

《基于主题模型和混合模型的微博客交叉话题发现的研究》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、西南交通大学硕士研究生学位论文第1I页AbstractMicro．blogisarelationshipbasedonuserinformationsharing，disseminationandaccesstotheplatform．Micro．bloghasbecomeoneofthemainsourcesofinformationontheIntemet．Itisverydifferentfromothernetworktext．Firstly,ithasrelativelysimplecontent(Itsmainbodyusuallyincludeslessthan140wo

2、rds)．Inaddition，itCanbepostedinreal．timebymobilephone，instantmessagingsoftwareandSOon，whichresultsinlargeamountsofdatainashortperiodoftime．Thiskindofdataisoftenhuge，messyandchaotic．Itisextremelydifficulttofindtheinterestinginformationaccuratelyandefficiently．Topicdetectiontechnologyisanewresearc

3、hfieldofnaturallanguageprocessing．Itfocusesonhelp堍theusercollectionandmergingoftheinformationthatdistributedunderthesametopic．Theusersfendtheinformationtheyinterestedinquicklyandaccurately．AlthoughthetraditionaltopicdetectionalgorithmsbasedonVSM(VectorSpaceModel)andclusteringalgorithmachievedgoo

4、dresultsandfacilitatedawiderangeofapplications，whendealingwithalargescaleMicro-blogshorttext，therearesomeshortcomings．Firstly，theyexisthigh．dimension，sparse，synonymyproblemswhenthedocumentsarepresentedbyfeamreveCtors．Inaddition，themostclusteringalgorithmsoftraditionaltopicsextractionarepartition

5、ingmethod，whichdidnotconsidertherelationshipbetweenthetopics，SOtherearesomelimitations．Underthesecireumstances，thetopicmodelisproposedastextrepresentationmodelaccordiIlgtothecharacteristicsofMicro．blog．Therearethreemaintopicmodels：LatentSem锄ticAnalysis(LSA)，ProbabilityLatentSemanticAnalysis(PLSA

6、)andLatentDirichietAllocation(LDA)．LDAisoneofthemostpopularandconlmonlyusedtopicmodelstoday，SOthisthesisutilizesLDAmodeltoextractthehiddenMicro-blogtopicsinformation丘omt：hedataset．Andthen，anoverlappingtopicdetectionalgorithmbasedonmixturemodelisproposedinordertosolvetheinsufficiencyoftraditional

7、topicdetectionalgorithm．Finally,aMicrobloggingoverlappingtopicdetectionsystemisestablished·Experimentalresultsonrealdatasetsshowthefeasibilityandvalidityofthealgorithm．Keywords：Microblog；topicmodel；overlappingtopicdetection；