基于内容相关性挖掘的反馈式搜索引擎框架  

Framework of Feedback Search Engine Based on Content Relevance Mining

作　　者：侯越先[1] 张鹏[1] 于瑞国[1] 

机构地区：[1]天津大学计算机科学与技术学院,天津300072

出　　处：《天津大学学报》2008年第8期941-945,共5页Journal of Tianjin University(Science and Technology)

基　　金：国家自然科学基金资助项目(60603027);天津市应用基础研究计划资助项目(05YFJMJC11700);微软亚洲研究院专项基金资助项目;海量科技基金资助项目

摘　　要：当前主流的搜索引擎根据查询词在网页中的出现频率,辅以网页权威性等信息,生成查询结果.但用户提供的查询词往往非常简单,因此搜索引擎难以确定用户的查询意图.为此,给出了一种利用海量clickthrough数据进行网页内容相关性挖掘的方法,在此基础上给出了一种反馈式搜索引擎(FSE)框架及相关算法.FSE根据网页相关性动态生成查询结果,以期提供给用户更中肯和个性化的信息.基于真实点击数据,进行了网页相关性矩阵的压缩实验和有效性实验,证明了该框架的可行性.Current mainstream search engines generate search results by analyzing statistical information such as the frequency of queries in web pages and the ranking of web pages. But search engines cannot determine what kind of information users want because queries are often simple in many situations. A web content relevance mining method was put forward which uses large amounts of clickthrough data. Furthermore, based on this method, a framework of feedback search engine （FSE）and associated algorithms were proposed. According to page-to-page relevance, FSE generated search results dynamically to provide users with more accurate and personalized information. Based on real clickthrough data, experiments on the compressibility and effectiveness of the web relevance matrix were performed. And the experimental results demonstrate the feasibility of the proposed framework.

关 键 词：WEB信息检索 反馈式搜索引擎 网页相关性 clickthrough数据 

分 类 号：TP393.4[自动化与计算机技术—计算机应用技术] TU375.4[自动化与计算机技术—计算机科学与技术]