Japanese-Chinese Phrase Alignment Exploiting Shared Chinese Characters

Common Chinese characters between Japanese and Chinese have been proved to be effective in Japanese-Chinese phrase alignment. Besides common Chinese characters, Japanese and Chinese also share many other semantically equivalent Chinese characters. However, there are no available resources for this kind of Chinese characters. In this paper, we propose a statistical method aiming to detect these Chinese characters which we call statistically equivalent Chinese characters. We exploit statistically equivalent Chinese characters together with common Chinese characters in a joint phrase alignment model. Experimental results show that our approach achieves over 1 point lower AER and 1 BLEU increase comparing to the baseline system.