A Method using Language Grid and Concept Base for Japanese- English Cross-language Information Retrieval

This paper describes query translation using language resources and a concept base method for Cross-language Information Retrieval (CLIR). In the proposed method, queries are translated by multiple machine translation systems on the Language Grid. The queries are then expanded by using a bilingual dictionary to translate compound words or word phrases. In addition, documents related to the translated query are retrieved with a TFIDF term weighting model. The top 100 retrieved documents are re-ranked by a specificity-considered concept base with the noun phrases and compound words extracted from the query. The reranked results are combined with the results retrieved by the probabilistic model. For evaluation of the proposed method, we use the average precision of the non-interpolated recall and precision to compare our method with the NTCIR1 participation systems. The proposed method achieved the highest precision.