一种基于局部分析面向事件的查询扩展方法

摘要

Aiming at the demand for event information by users, we propose a method--loeal analysis-based event-oriented (LA-EO) query expansion, whieh divides query terms into two eategories: event terms and qualifying terms. The paper emphatieally diseusses the analysis of event-oriented query terms, the expansion of event terms and the eomputing between query terms and doeuments. We have eolleeted 4011 doeuments and set ten query topies eentering on emergeney domain to evaluate our proposed methods. The results show that LA-EO query expansion offers more effeetive performanees for retrieving events, eompared with LA-Roeehio and LA-LCA.%针对用户获取事件类信息的需求,提出了一种基于局部分析面向事件LA-EO(local analysis-event-oriented)的查询扩展方法,该方法将查询项区分为事件项和限定项两类分别处理.文章重点讨论了面向事件的查询项分析、事件项的扩展以及查询项与文本相似度的计算等问题.围绕突发事件领域,使用搜索引擎和定点采集相结合的方法收集了4011篇文本,设置了10个查询项对本文提出的方法进行了实验比较.结果表明:LA-EO与Rocchio机制(记作LA-Rocchio)和局部上下文分析(记作LA-LCA)扩展方法相比,对事件类信息的检索,LA-EO具有更优的检索性能.