Utilizing Images for Assisting Cross-Language Information Retrieval on the Web

In this paper, we discuss the effectiveness of employing visual clues in interactive cross-language information retrieval to assist users in selecting appropriate translation candidates for a query term. Experimental results show that by associating images collected by a Web image search, for each translation candidate, the selection of correct translations was considerably improved (by 29.3%) and the efficiency of the selection was enhanced, i.e., the elapsed time in selection was substantially reduced (by 16.7%) compared to the cases where only back-translated terms were presented. The results suggest that visual items, even if simply collected by a Web image search without any image sense disambiguation, can be superior to verbal information as clues for query term selection in target languages that the user is not familiar with.