Statistical Machine Translation between Myanmar and Myeik

Abstract

This paper contributes the first evaluation of the quality of machine translation between Myanmar and Myeik (also known as Beik) . We also developed a Myanmar-Myeik parallel corpus (around 10K sentences) based on the Myanmar language of ASEAN MT corpus. In addition, two types of segmentation were studied word and syllable segmentation. The 10 folds cross-validation experiments were carried out using three different statistical machine translation approaches: phrase-based, hierarchical phrasebased, and the operation sequence model (OSM). The results show that all three statistical machine translation approaches give higher and comparable BLEU and RIBES scores for both Myanmar to Myeik and Myeik to Myanmar machine translations. OSM approach achieved the highest BLEU and RIBES scores among three approaches. We also found that syllable segmentation is appropriate for translation quality comparing with word level segmentation results.