English-Vietnamese Cross-Language Information Retrieval: An Experimental Study

摘要

Translation units play an important role in Cross-Language Information Retrieval (CLIR) involved with languages lacking word boundary - such as Chinese and Vietnamese. While the impact of these units has been well studied in English-Chinese CLIR, it has not been considered in English-Vietnamese CLIR. Therefore, in this paper, we examine how different translation units contribute to the English-Vietnamese CLIR performance. Similar to translation units, different translation resources tend to have different effects on CLIR. Thus, the use of these linguistic resources should also be investigated. Since bilingual dictionary is the only resource available, we have constructed an English-Vietnamese parallel corpus using automatic system based on web mining. We then experiment three methods to exploit these resources - using the dictionary only, the parallel corpus only, and the parallel corpus to disambiguate translations given by the dictionary. Our experiments have shown that automatically generated parallel corpus is very feasible for both translation and disambiguation in English-Vietnamese CLIR.