A new similarity computing method based on concept similarity in Chinese text processing.pdfVIP

  • 16
  • 0
  • 约3.15万字
  • 约 16页
  • 2018-04-10 发布于河南
  • 举报

A new similarity computing method based on concept similarity in Chinese text processing.pdf

A new similarity computing method based on concept similarity in Chinese text processing.pdf

ScencefnChnnS已rf8sF ,n厂0丌 fDn cfence wwW.sCiChina.COm infO.SCiChina.cOm @ 2008 SClENCEINCHINAPRESS www.springer¨nk.cOm Sp一ringer Anew simiIaritycOmputingmethOdbasedOn CcOOnnCCeeDDttSslimmliIIaarri1ttVVi1nnC乙hni1nneeSseetteeXxttDDrrOOCceeSsSsi1nngg PENGJjng’,2十YANGDongQing,TANGShiWei’WANGTengJiaO’&GAOJun’ , , ’SchooIofEIectrDnicsEngineeringandComputerScience,PekingUniVersjty,Be_jing1O0871,China; 。Department0fScienceand。rechnoIogy ChengduMunicipaIPubIicSecurity,Bureau,Chengdu61O017 , China ThepaperprOposesanew textsimiIarity computingmethodbasedon conCept sim llarityjnChinesetextprOcessing.Thenew methOdconvertstexttOwOrdsVec· tOrspacemOdeIatfirst,and then spIitswOrds into a setOfcOncepts.ThrOugh cOmputingtheinnerproductsbetweenconcepts,itObtainsthesim_lar-tybetween wOrds.The new methodcOmputesthesim_larityOftextbased OnthesimiIarity0f wordsatlast.ThecontributionsofthepaperincIude:1 proposeanewcomputing formulabetweenwords;2 proposeanew te】ctsim_laritycompUtingmethodbased onwordssim-larity;3 successfu¨yusethemethodintheapplicationofsimiIarity computingofWEBnews;and4 provethevaIidityofthemethodthroughextensiVe experiments. concep1similar ,simIlaritycompung,Vec1orspace,|nnerpr0ductspace 1 IntrOduCti0n Thetextinfom atiOnretrieValisanimpOrtantpartintheinfomlatiOntechnOlOgy’andtexts1mllar— itycomputingisOneofthehOt—spOtresearchfieldsintextinf0肌 atiOnsearches.ThepurpOse0f sirnjlaritycomputingpr0posedinthispaperistocOmparethesimilardegreebetweenanypairsof textstrings.ThiscOmputingmethOdcanbewidelyusedintheareaOfnaturallanguageprOcess, such astextautOmatica1classificatiOn,c1ustering,fuzzy query,machinetmnslation,andautO— matictab1oid. ThetypicalmOdelsintextsimilaritycomputinginc1udeBOOleanm0de1,VectOrspacemOde1 VSM ,andpr0babilisticmode1.ThesemodelsrespectiVe1yusedi仟erentmethodst0pr0cess characterweight,classmablestudy,andtheVSM isoneofthemostef-fectivemodels. ReceivedAugust28,2oo7;d|cceptedJanuarv1l,2008 d0i:l0.1OO7/sl1432一o08.0103—4 十corresponding

文档评论(0)

1亿VIP精品文档

相关文档