潜在语义分类模型的中文分类性能研究.pdf

计算机研究与发展 ISSN1000—1239／CN11—1777／TP of Researchand 42(增刊)：367～372，2005 JournalComputer Development 潜在语义分类模型的中文分类性能研究 曾雪强1 王明文2 陈素芬3 1(江西医学院计算机教研室南昌330006) 2(江西师范大学计算机信息工程学院南昌330027) 3(南昌工程学院计算中心南昌330099) (zxq@jxmu．edu．an) PerformanceofaLatentSemanticClassificationModelonChinese Analysis Corpus ChenSufen3 Mingwen2，and ZengXueqian91，Wang andResearch Medical 330006) 1(ComputerTeaching Section，JiangxiCollege，Nanchang Normal 330027) 2(SchoolofComputerInformationEngineering，JiangxiUniversity，Nanchang Institute Cozier，NanchangofTechnology，Nanchang330099) 3(Computer iSanewclassificationmodelwhichextendslatent AbstractLatentsemanticclassification(LSC)model semantic theclassificationinformationofthe documentinto indexing(LSI)modelbyintroducing training thelatentsemanticstructureviaasecondsetoflatentvariables．TheLSCmodelcanovercomethe problem be becauseofthesmallvalues intheLSImodelthatsome featuresforclassification important mayignored oftheir this effectofthe￡factorintheLSCmodelis correspondingeigenvalue．Inpaper，the investigated andsome resultsareobtained．The differencebetweentheLSCmodelandother interesting performance classificationmodelsisalsostudied，andthe showthattheLSCmodel betterthanthe experiments performs classificationmethodssuchaskNNandSVM． existing wordstext semantic semantic