基于查询扩展和数据融合的检索过程优化

Optimization of Retrieval Procedure Based on Query Expansion and Data Fusion

介绍典型的检索过程优化方法——数据融合和基于相关度反馈的查询扩展,前者通过集成多个检索结果提高检索性能,后者执行多次查询,依据前次结果修改/扩展用户查询,以求更好地反映用户信息需求,并在此基础上提出一种新的检索过程优化方法——HQD方法,由相关度反馈结果生成多个替代查询,在检索这些替代查询后,采用求和余弦法生成最终检索结果。仿真实验结果表明,该方法是有效的。

Typical optimized retrieval procedure methods such as data fusion and relevance feedback-based query expansion are introduced. While data fusion improves retrieval performance by merging multiple retrieval results, relevance feedback revises user query according to previous retrieval result and runs the new query to improve retrieval performance. On the basis of this, a novel optimized retrieval procedure method called HQD is presented, which selects top-ranked documents from the relevance feedback result, runs these documents as surrogate queries, and merges the retrieval results using a sum cosine measure. Experimental results show this method is effective.