A New Approach for Multi-Document Update Summarization-一种多文档更新摘要的新方法.pdfVIP

XYeta1．Anew for multi—documentsummarization．JOURNALOF update LongC，HuangML，Zhu approach SCIENCEANDTECHNOLOGY 2010．DOI 25(4)：739-749July 10．1007／s11390-010-1057-8 ANew forMulti—DocumentSummarization Approach Update Huan91(黄民烈)，Xiao-YanZhul，+(朱小燕)，Member,CCF ChongLon91(龙狮)，Min—Lie and MingLi2(李明)，Fellow,ACM,IEEE 1State and National Science KeyLaboratoryo／IntelligentTechnologySystems,TsinghuaLaboratory，DrInformation Scienceand 1 and of Technology，Tsinghua 00084，China Technology,DepartmentComputer University，Bering WaterlooN2L Canada 2School WaterlooI 3GII o|ComputerScience．Universityol E-maih ReceivedOctober 22，2009；revised8，2010． April AbstractFast ontheInternetcanbe more withthe ofautomaticdocument changingknowledge acquiredefficientlyhelp and describesanovel formulti-document summarization updatingtechniques．Thispaper approach update Thebest isdefinedtobetheonewhichhastheminimuminformationdistancetotheentiredocumentset． summary Thebest hastheminimumconditionalinformationdistancetoadocumentclusterthata updatesummary given prior documentclusterhas been onthe 2007to2009 alreadyread．ExperimentsDUC／TAC have thatourmethod correlateswiththehumansummariesand http：／／www．nist．gov／tac／1proved closely outperforms LexRank undertheROUGE criterion． other suchas