A transitive translation for Indonesian-Japanese CLQA (自然言語処理) A transitive translation for Indonesian-Japanese CLQA

This paper investigates a Cross Language Question Answering for a language pair with limited resources. In order to solve the limited translation resource problem, we use a transitive translation with English as the pivot language. To select the best translation, we use a mutual information score calculated from the corpus of the target language (e.g. Japanese). As an additional advantage, we build the CLQA using some machine learning based modules. By this, the system can be easily adapted into other language pair. In the experiments, we use test set of NTCIR 2005 CLQA1 task, consists of 200 questions. We also compare the transitive translation result with a direct translation result which employs a middle size Indonesian-Japanese dictionary.