Description of the NTU Japanese-English Cross-Lingual Information Retrieval System.

Abstract

This paper describes a Japanese-English Cross-Language Information Retrieval (CLIR) System for the evaluation in NTCIR project. We extend our work on Chinese-English CLIR to deal with this porblem Sevewal query translation strategies,including select-all,select-first,and co-occurrence models by different corpora,and several query generation strategies,including topic,description,narrative,and concept fields,are considered. The experiment results show that (1) the combination of topic and concept fields outperform other combination; and (2) select-first is the best,co-occurrence by TREC-6 and LOB,and select-all is the worse in EOI task. The evaluation in EO2task is similar Co-occurrence by TREC6 and select-first on topic and concept combination rank I and 2, respectively.