文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Text Data Mining Theory and Methods
侵权投诉

Text Data Mining Theory and Methods

This paper provides the reader with a very brief introduction to some of the theory and methods of text data mining. The intent of this article is to introduce the reader to some of the current methodologies that are employed within this discipline area wh

Je Solka/TextDataMining:TheoryandMethods95

1.Introduction

AlvinTo er[52],thefamedauthorofFutureShockandother neworks,onceproclaimed:“Knowledgeisthemostdemocraticsourceofpower.”Ibelievethisideaholdsgreatweightasweproceedforwardintothe21stcentury.Anumberofcompanieshavecreatede ectivebusinessmodelsbasedonthemanagementofinformation.Inthesetimes,theabilitytoextractinformationfrommanyanddisparatesourcesofdatawillhelpdetermine,inpart,thebalanceofpowerbetweencorporationsandamongnations.

Besidestheusualquantitativeorqualitativedatafamiliartostatisticians,thedatathatservesasinputtotheinformationextractionalgorithmscantakeanyform,includingimagery,video,audioortext.HereIwillfocusontext-baseddatasources.Thetextualdatasourcesforinformationextractioncanrangefromfreeformtexttosemi-formattedtext(html,xmlandothers)andincludesthosesourcesthatareencodedinopensourcedocumentformats(OpenDocument)andotherproprietaryformats(MicrosoftWordandMicrosoftPowerpoint).Ifeelthattheproblemofextractinginformationfromthesedatasourcesisonethato ersgreatchallengestothestatisticalcommunity.Theplanforthisarticleistodiscusssomeofthesechallengesandtogenerateinterestinthecommunityinthisparticulartopicarea.

Dataminingontexthasbeendesignatedatvarioustimesasstatisticaltextprocessing,knowledgediscoveryintext,intelligenttextanalysis,ornaturallan-guageprocessing,dependingontheapplicationandthemethodologythatisused[1].Examplesoftextminingtasksincludeclassifyingdocumentsintoasetofspeci edtopicareas(supervisedlearning),groupingdocumentssuchthateachmemberofeachgrouphassimilarmeaning(clusteringorunsupervisedlearning),and ndingdocumentsthatsatisfysomesearchcriteria(informationretrieval).Intheinterestofspace,Iwillnotprovidediscussionsoninforma-tionretrieval.ThereaderisreferredtotherecentworkbyMichaelBerrythatprovidesdiscussionsonsomeoftherecentworkintheseareas[5].

Iplantodescribethetextminingproblemwithintheframeworkofthosestepsthatneedtobeperformedinvirtuallyanydataminingproblem.The rststepistheextractionoffeaturesonthedocumentcollectionssothatonecanperformcomputationsandapplystatisticalmethodologies.Theextractedfeaturesshouldsomehowcapturethecontentofthedocuments.Ideally,theyshouldcapturethecontentinsuchamannerthatdocumentsdiscussingsimilartopics,butwithdi erentterminology,wouldhavesimilarfeatures.

Next,onemustformulateawaytomeasurethedistancesbetweenthedoc-uments,wherethedistanceisconstructedtoindicatehowsimilartheyareincontent.Aswillbeshown,thisnotonlyallowsustoapplyclassi cationandclusteringmethods,italsoenablesonetoformulatestrategiestoreducetheinherentdimensionalityoftheencodeddocumentfeatures.(Thepurposeofdi-mensionalityreductionistoprovideamoremeaningfulgeometrizationofthefeatures.)Givenarenderingofthefeaturesinalowerdimensionalspace,wecanthenturnourattentiontoclustering,discriminantanalysis,orvisualizationofthedocumentinformationspace.Thesubsequentsectionsofthisarticleserveto

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email