基于查询日志的局部共现查询扩展

QUERY EXPANSION OF LOCAL CO-OCCURRENCE BASED ON QUERY LOG

查询扩展是信息检索中的一个关键问题,查询扩展的有效性决定了检索系统的检索性能。大多数的查询扩展基于全局分析或者局部分析法,虽然准确率有了很大的提高,但是都有各自的局限性。查询日志是大量用户长期查询行为的记录。提出了基于查询日志的局部共现查询扩展方法,该方法通过挖掘用户初始查询与查询日志之间的联系,构建一个用户初始查询与用户文档的关联关系图,并且使用局部共现的方法构建相关词表,从而实现查询扩展。在50 000篇语料上的测试表明,该方法相对于未扩展时准确率提高了44%以上。

Query extension is a key issue in information retrieval, the efficiency of query expansion determines the retrieval performance of retrieval system. Most of the query expansions are based on global analysis or local analysis, though the accuracies have been greatly improved, but they all have their own limitations. Query log is the record of long term query behaviour by a great quantity of users. In this paper, we propose a query log-based expansion method of local co-occurrence, through which we can build an associated diagram of user initial query and user document through mining the link between user＇s initial query and user logs, and construct the related word list using local co-occurrence method, thus to realise the query expansion. The test on 50,000 corpora shows that the precision has about 44% improvement after using this method.