Improving Retrieval Performance by Global Analysis

In this paper, we propose a global analysis approach for query expansion in information retrieval. This approach, which concentrates on extracting associated expansion terms for a given query so as to improve the retrieval performance, can be divided into three stages. The first stage, called association calculation, is to measure associations between term-term pairs. The second stage focuses on how to select suitable terms for expansion. In the last stage, a reweighting scheme is introduced in order to clarify the impact of each expansion term. Experimental results on the TREC collection demonstrate the effectiveness of our approach