Improved Query Language Model Generation Using Mixture Models

摘要：

This paper presents a novel method of using mixture models with document segmentation to improve the performance of pseudo relevance feedback for document retrieval.Pseudo-relevance feedback in information retrieval traditionally selects a number of topranked documents,from which terms highly related to the initial query are selected to form a new query model for query expansion.However,the query expansion often causes query shift since not all the contents within the document are coherently relevant.In this paper,we propose to build mixture models by combining the original query information with the expanded query language model using aspect model.The experimental results have shown that our proposed methods significantly outperform a number of strong baselines on several large scale TREC collections.

作者: Qiang Huang Yong Huang HongYan Yi

作者单位: School of Computing Sciences University of East Anglia Norwich,UK School of Information Engineering HeNan Institute of Science and Technology XinXiang,China

会议类型: 国际会议

会议名称: 2010 4th International Conference on Intelligent Information Techonlogy Application(第四届智能信息技术应用国际学术研讨会 IITA 2010)

会议地点: 秦皇岛

会议语种:英文

页码: 8-11

在线出版日期: 2010-11-05（万方平台首次上网日期，不代表论文的发表时间）