Large Scale Myanmar to English Neural Machine Translation System

Myanmar language is one of the low resource languages. There are a few resources that include bilingual sentences. Therefore, there are some difficulties to collect or crawl the parallel sentences. Existing Myanmar Translation systems use rule-based as well as statistical-based approach with the small amount of parallel corpus. Therefore, the performance of Myanmar to English Machine Translation system is still low. In this paper, large scale parallel corpus is prepared and introduces the Myanmar to English Neural Machine Translation system. Nowadays, neural machine translation models became a popular research field and it reaches good results in some languages. In this work, we did the experiment on the word-level model and character-level model based on neural method for Myanmar to English translation. The evaluation results show that neural machine translation models lead to improve the performance of Myanmar to English translation.