基于网络日志的用户兴趣模型构建

Web-Query-Log-Based User Interest Model

了解用户查询意图对改善搜索引擎质量起到了至关重要的作用,对具有特定兴趣的用户进行查询分析,使搜索引擎更能了解用户的真实需求。本文通过对网络查询日志进行聚类分析,将相似度大的查询词聚类,建立用户兴趣模型对用户的兴趣进行分析。根据查询词内容重合度,建立查询词图,并结合查询词的PageRank算法,提出一种基于用户查询词概率分布的评价方法,对用户感兴趣的查询词进行评价。最后,根据查询词的概率分布将最感兴趣的查询词推荐给用户。

The understanding of the intend of user queries plays a crucial role,and we analyze the user with specific insterest so that the search engines can better understand the real needs of users.By the que ry cluster of network query log in the paper,we piece together user queries with the largest similarity,and build user interest model for the analysis of the user interest.we set up query term map by the overlap ra tio of query content,and put forward a evaluation method based on the probability distribution of queries so that we evaluate queries with ther user interest.Finally,According to the probability distribution of que ries,we recommend the most interested queries to users.