资源描述：
《微博短文本检索关键技术研究》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、哈尔滨工业大学工程硕士学位论文4.基于排序学习模型的微博检索。在本章中我们采用了一种新的排序策略，在排序中不仅考虑了文本相似度，而且同时考虑了微博用户和微博独有的一些特征信息。因此采用了排序学习方法，融合多个特征，来产生更好的排序函数。进行了一系列实验，通过对单个特征的分析和尝试不同的特征子集选择方案，来获取能达最佳微博检索性能的特征子集。关键词：微博检索；语言模型；时间信息检索；参考文档模型；排序学习-II-哈尔滨工业大学工程硕士学位论文AbstractAmicoblogisashorttextme
2、ssagesofnomorethan140characterssharedthroughasocialmedia,suchasTwitterorSinaweibo.Microblogs,usedbyhundredsmillionsofusers,appearattractiveinsofarastheypromiseaccesstotimelyinformationwrittenbypeoplewhotheyhavechosentopayattentionto.Withinthepastfiveyea
3、rs,microblogshavegreatlydevelopedandnowbecomeatypicalrepresentativeofsocialmedia,andalsooneindispensabletimelyinformationsource.Inthiswork,weusemicroblogtodenotethosemicroblogssharedthroughTwitter.Microblogdatahasgrownexplosively.Thus,howtohelpuserexact
4、lyfindmicroblogpoststhattheyareinterestedbecomesanimportantmissionofthemicrobloginformationretrieval.Owingtothelimitedtextsize，informallywritten,noisyandreal-timenatureofmicroblogs,traditionalIRmodelsrunintodifficulty.Inordertoaddressthisissue,thispaper
5、researchesonthekeytechnologiestoresolvethepoignantproblemsfacedinmicroblogretrieval.Indetails,thisthesisiscarriedoutinrespecttothefollowingfouraspects:1.Hottimebasedlanguagemodelingapproach.Inthissection,wefirstinvestigatetwomethodsoftimebasedlanguagemo
6、delwhichbothareunderthehypothesisof“thenewerdocumentis,themoreimportantitis”.Thenwedemonstratethatthisassumptionmaynotalwaysworkbytheanalysisofrelevancedocument’stemporaldistributionforspecificquery.Finally,wedefinequery’shottime,andproposedahottime-bas
7、edlanguagemodelapproachtoretrievalmicroblogposts.Meanwhile,wedoanexperimentalcomparisontopriorapproaches.2.QueryModelingintegrationwithtemporalinformation.Inthissection,weutilizetemporalproperties(e.g.recencyandtemporalvariation)toenrichauserquerytoimpr
8、overetrievalperformance.Indetail,weexplorethreequeryexpansion(QE)methods.ThefirstQEmodelbaseduponrecencycansuggestcandidatetermsforquerywhichfavorsmorerecentdocument.ThesecondQEmodelcanhandletemporalvariationsconsistingofanoldpea
