Unsupervised bilingual segmentation using MDL for machine translation

In statistical machine translation systems, a problem arises from the weak performance in alignment due to differences in word form or granularity across different languages. To address this problem, in this paper, we propose a unsupervised bilingual segmentation method using the minimum description length (MDL) principle. Our work aims at improving translation quality using a proper segmentation model (lexicon). For generating bilingual lexica, we implement a heuristic and iterative algorithm. Each entry in this bilingual lexicon is required to hold a proper length and the ability to fit the data well. The results show that this bilingual segmentation significantly improved the translation quality on the Chinese-Japanese and Japanese-Chinese sub-tasks.