IMPROVE CROSS LANGUAGE INFORMATION RETRIEVAL WITH PSEUDO-RELEVANCE FEEDBACK

Abstract

In dictionary-based Cross-language Information Retrieval systems, structured query translation has been shown to be an useful method for improving system performance. In this paper, we examine the effects of using pseudo relevance feedback to refine the structured query in the target language. We propose different methods for term weighting based on word distributions and the mutual information between expanded terms and original query terms. Our experimental results in a dictionary-based Vietnamese-English CLIR system show that while changing query terms weights has effects on improving precision, query expansion improves recall rates. The combination of these two techniques helps to improve system performance up to 12%, in terms of Mean Average Precision.