The Cross-Language Query Expansion Algorithm Based on Hybrid Clustering

Abstract

To propose a cross-language query expansion algorithm based on hybrid clustering aiming at problems such as theme shift, word dismatch, translation ambiguity and polysemy etc. in the cross-language information retrieval. The HPH-CLQE for cross-language query expansion is put forward by combining advantages of partition clustering and hierarchical clustering. This algorithm can achieve a good clustering effect by avoiding selection of the initial cluster center and cluster number. Then, the CLQE-HPH based on hybrid clustering is presented. This algorithm applies HPH-CLQE to conduct clustering analysis on top documents selected at the initial cross-language retrieval aiming to obtain the expansion word that is highly relevant with user query to achieve the cross-language query post-translation expansion. The experimental results based on NTCIR-5 CLIR data set indicate that MAP, P@5 and P@10 related to the algorithm proposed in this paper get improved greatly with the maximal increase of 20.52% by comparing with three standard algorithms. This algorithm can reduce problems such as theme shift and word dismatch etc. in the cross-language information retrieval effectively and improve the cross-language information retrieval performance.