Context-based query expansion method for short queries using latent semantic analyses

Abstract

Short queries are the key difficulty in information retrieval (IR). A plenty of query expansion techniques has been proposed to solve this problem. In this paper, we propose three different models for query suggestion using the cosine similarity (CS), the Language Models (LM) or their fusion. The expansion terms are selected using the Latent Semantic Analyses method based on the result of the three query suggestion methods. The approaches proposed improve the precision of the user query by adding additional context to it. Experimental results show that expanding short queries by our approaches improves the effectiveness of the IR system by 48,1% using the CS based model, 19,2% using the LM model, and 13,5% using the fusion model.