查询语义图辅助的信息检索性能预测模型  

Predicting Query Performance Using Semantic Chart

作　　者：乔亚男[1] 齐勇[1] 

机构地区：[1]西安交通大学电子与信息工程学院,陕西西安710049

出　　处：《电子学报》2011年第A03期158-162,共5页Acta Electronica Sinica

基　　金：国家自然科学基金重点项目(No.60933003)

摘　　要：查询性能预测技术试图在进行费时的实际信息检索之前对特定查询的性能进行预测,以便根据预测结果在不影响查询所代表的信息需求的基础上对查询进行调整,提高最终检索结果的精确度.针对传统查询性能预测模型没有考虑查询词间语义关系的问题,本文提出了一种查询语义图辅助的信息检索性能预测模型,该模型将表征查询词间语义关系的查询语义图引入性能预测的过程中,使查询性能预测模型避免了查询词独立性假设.实验结果表明,经过查询语义图加权的性能预测模型的预测精确度明显高于传统的性能预测模型,预测结果与实际检索结果的相关度最高提升了约46.32%.The technologies of predicting query performance try to predict the performances of queries before actual time-consuming information retrievals,and adjust the original queries based on the predicting results to improve the performances without modifying the information needs.For avoiding the terms-independence assumption in traditional query performance predicting models,this paper proposes a query performance predicting model using semantic charts of queries which describe the semantic relations between query terms.The experimental results show that the precisions of semantic charts weighted query performance predicting models are higher than traditional models significantly,and the relativities between predicting performances and actual performances are improved about 46.32% at most.

关 键 词：信息检索 查询性能预测 查询语义图 核心词 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]