欢迎来到天天文库
浏览记录
ID:19726561
大小:500.50 KB
页数:36页
时间:2018-10-05
《信息过滤(information filtering)综述》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、信息过滤(InformationFiltering,IF)综述中科院计算所软件室王斌wangbin@ict.ac.cn2001.12.10主要内容IF的基本概念IF系统的分类IF系统的组成IF系统的评估IF的现状及发展趋势一、基本概念定义IF定义:从动态的信息流中将满足用户兴趣的信息挑选出来,用户的兴趣一般在较长一段时间内不会改变(静态)。SelectiveDisseminationofInformation(SDI),来自图书馆领域。Routing,来自MessageUnderstanding。Curre
2、ntAwareness,DataMiningIFvsIR/分类/IEIF&IR:广义地讲,IF是IR的一部分Database动态,需求静态;Database静态,需求静态UserProfilevsQueryIF用户要对系统有所了解,IR不需要。IF要涉及到用户建模/个人隐私等社会问题IF&CategorizationCategorization中的Category不会经常改变。相对而言,UserProfile会动态变化IF&IEIF关心相关性,IE只关心抽取的那些部分,不管相关性IFapplications
3、InternetSearchResultsFilterPersonalEmailFilterListServer/NewsgroupFilterBrowserFilterFilterforchildrenFilterforcustomers:recommendation二、IF分类体系IF分类示意图InitiativeofoperationActiveIFsystemsCollectandsendrelevantinfotousersPushtousersInfooverload,somakeaccurat
4、euserprofilePassiveIFsystemsNotcollectinfoforusersEmailorUsenetnewsLocationofoperationAttheinfosourcePostprofilestoinfoproviderClippingserviceUsuallypayfeeAtafilteringserverInfoprovidersendinfotoserverServedistributedinfotousersAttheusersiteLocalfilterings
5、ystemSuchasoutlook&NetscapeEmail&FoxmailFilteringapproachCognitivefilteringContent-basedfilteringDocumentcontentvsuserprofilesSociologicalfilteringCollaborativefiltering,orproperties-basedfilteringSimilaritybetweenusersRecommendationsystemsUsermodeling&Use
6、rclusteringComplementforcontent-basedsystemsMethodsofacquiringknowledgeaboutusersExplicitapproachUserinterrogationFillingformsImplicitapproachRecordinguserbehaviorTime/times/context/activity(save/discard/print/browsing/click)/etc.Explicit&ImplicitapproachD
7、ocumentspace(case-based)Stereotypicinference(predefineddefaultprofile,thenchangeduringscanning)三、IF系统的组成一般组成(d)LearningComponentUserInformationProvider(b)FilteringComponent(a)DataAnalyzerComponent(c)User-ModelComponentupdatesfeedbackrelevantdataitemsrepres
8、enteddataitemsdataitemspersonaldetailsuserprofileData-analyzercomponentBeclosetotheinfoproviderObtainorcollectdatafromtheinfoproviderAnalyze&representdocuments(suchasBooleanModel,VSM,etc)Passtherepresentation
此文档下载收益归作者所有