Query expansion method based on word contribution (poster abstract)

Query expansion (QE) has been considered as one of the most indispensable methods to achieve successful information retrieval. The application of QE lightens the burden imposed on the user, who would otherwise need to generate an effective query by themselves. One of the widely known methods of QE is the method based on Rocchio’s algorithm[2], which is based on the vector space model. While QE based on Rocchio’s algorithm has been proved to achieve excellent performance[l], we believe an even more effective QE can be achieved by applying the discriminativeness of words extracted from relevant documents. This proposed weighting scheme can be implemented by extracting not only the importance of the word in relevant documents, but also the influence of the word to query-document similarity. In this paper, we propose a novel QE method based on a measure called word contribution, and prove its effectiveness by experiments.