Improving Statistical Myanmar to English Machine Translation through Morphological Analysis

Abstract

In this paper, we present a translation model of Myanmar phrases for statistical machine translation. Phrase based SMT models have limitations in mapping from the source to target language without using linguistic information. Morphological analysis is needed especially for morphology rich language and small amount of training data are available. Myanmar language is inflected language and we have very modest parallel resources for machine translation. Therefore, we present Myanmar language morphology analysis in noun and verb phrases. Especially analysis is performing on number category of noun phrase, suffixes and tense particle of verb phrases. We test our system on Myanmar-English bilingual corpus. The experiment results show that the quality of statistical machine translation is improved by applying morphology analysis of Myanmar language.