资源描述：
《detect hot topic aging theory.pdf》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、HotTopicDetectiononBBSUsingAgingTheoryDonghuiZhengandFangLiUDS-SJTUJointResearchLabforLanguageTechnologyDept.ofComputerScienceandEngineeringShanghaiJiaotongUniversity,Shanghai,Chinazhdhui@gmail.com,fli@sjtu.edu.cnAbstract.BBS(BulletinBoardSystems)isoneofthemostcommonplacesforthreadeddiscussion.Itbe
2、comesmoreandmorepopularamongwebusers,especiallyinChina.Everydayahugeamountofnewdiscus-sionsaregeneratedonBBS.Itistoodifficulttofindhottopics.Tosolvethisissue,weproposeanovelapproachtodetecthottopicsonBBSforanyperiodoftime.Oursolutionconsistsofthreesteps.Firstofall,can-didatetopicsareextractedusingthecl
3、usteringmethod.Secondly,basedontheextractedtopics,agingtheoryisemployedtovaluatethehot-nessoftopics.Bothtwostepsabovearecarriedoutincrementallyovertime.Finally,topicsarerankedandhottopicsaredetected.ExperimentsperformedonpracticalBBSdatashowthatourmethodisquiteeffective.Keywords:Hottopicdetection,Agi
4、ngtheory,BBS.1IntroductionAsReadWriteWeb1indicates,therearenearly2hundredmillionofBBSusersinearly2008inChina.ThetotalnumberofdailypageviewsacrossBBShasreachedover1.6billion,with10millionpostspublishedeveryday.However,toomanytopicsarediscussedonBBS.HowtoidentifyhottopicsonBBSbecomesmoreandmoreimporta
5、nt.HerearesomebasicconceptsforBBS:Post:Apostisausersubmittedmessage.Thefirstpoststartsthethread;thismaybecalledtheentry.Poststhatfollowinthethreadaremeanttocontinuethediscussionabouttheentry,orrespondtootherreplies.Apostmainlycontainsfourparts:theauthor,thetitle,thetime-stampandthecontent.Thread:Athr
6、eadisacollectionofposts.Eachpostbelongstoandonlybelongstoonethread.Athreadcancontainanynumberofposts,includingmultiplepostsfromthesamemember,eveniftheyareoneaftertheother.Topic:AtopiconBBSisformedwithoneormorethreads.SupportedbytheNationalScienceFoundationofChinaunderGrantNo.60873134.1http://www.re
7、adwriteweb.com/W.Liuetal.(Eds.):WISM2009,LNCS5854,pp.129–138,2009.cSpringer-VerlagBerlinHeidelberg2009130D.ZhengandF.LiNowadays,togethottopicsonBBS,themostpopularmethodistoseeTop-Ntopicsbasedonthenum
