Abstract: The traditional local context analysis approach has the limitation of the high dependence on initial search results. To solve this problem, this paper gets the users’ most interested articles through statistical analysis and filtration of the user query logs, to replace the initial N articles, as the source of the query expansion words. The relevance between the query words and the articles’ words is counted using the local context analysis approach. The experiments show that this method has effectively improved the precision of the query.

Key words: local context analysis, user query log, query expansion, filtration

摘要： 传统的局部上下文分析其应用效果高度依赖于初次检索的结果。针对此局限，通过对用户查询日志的统计分析和过滤，得到用户最可能感兴趣的文章，代替初始检索得到的N篇文章，作为查询扩展词来源文档集，用局部上下文分析方法计算词间相关度。实验结果表明，该方法能够较大地提高查询精度。

关键词: 局部上下文分析, 用户查询日志, 查询扩展, 过滤