English-persian text retrieval using concept graph

Abstract

Cross-language information retrieval (CLIR) is the retrieval process where the user presents queries in one language to retrieve documents in another language. In this field the resolution of lexical ambiguity in translating queries is a key challenge. In this paper, we propose a technique for calculating translation probabilities based on creating query terms' concept graphs for selecting the right translation sense of query terms for English-Persian text retrieval. We present an efficient statistical method for creating this graph. We test the effectiveness of the proposed disambiguation method on Hamshahri collection1 that is standardized according to CLEF standards. Evaluation using this data collection shows great effectiveness of the proposed method.