Japanese-Chinese Phrase Alignment Using Common Chinese Characters Information

We describe a method to detect common Chinese characters between Japanese and Chinese automatically by means of freely available resources and verify the effectiveness of the detecting method. We use a joint phrase alignment model on dependency trees and report results of experiments aimed at improving the alignment quality between Japanese and Chinese by incorporating the common Chinese characters information detected by proposed detecting method into the alignment model. Experimental results of JapaneseChinese phrase alignment show that our approach could achieve 0.73 points lower AER than the baseline system.