Myanmar Phrases Translation Model with Morphological Analysis for Statistical Myanmar to English Translation System

Myanmar Phrases Translation Model with Morphological Analysis for Statistical Myanmar to English Translation System Thet Thet Zin a, Khin Mar Soe b, and Ni Lar Thein a a University of Computer Studies,Yangon,Myanmar thetthetzin.ucsy@gmail.com bNatural Language Processing Laboratory University of Computer Studies, Yangon, Myanmar Abstract. This paper presents Myanmar phrases translation model with morphological analysis. The system is based on statistical approach. In statistical machine translation, large amount of information is needed to guide the translation process. When small amount of training data is available, morphological analysis is needed especially for morphology rich language. Myanmar language is inflected language and there are very few creations and researches of corpora in Myanmar, comparing to other language such as English, French, and Czech etc. Therefore, Myanmar phrases translation model is based on syntactic structure and morphology of Myanmar language. Bayes rule is also used to reformulate the translation probability of phrase pairs. Experiment results showed that proposed system can improve translation quality by applying morphological analysis on Myanmar language. Keywords: Morphological Analysis, Statistical Machine Translation, Bayes rule, Syntactic Structure. 1 Introduction Machine translation (MT) is the task of automatically translating a text from one natural language into another. There exist different approaches to address the problem of machine translation. This paper presents translation model of Myanmar phrases for statistical Myanmar to English machine translation system. Target and source language model based on N-gram (trigram) and translation model based on Bayes’ rule to reformulate translation probability P (f| e). N-gram method (trigram) based source language model is used to extract phrases for segmented Myanmar sentence. Myanmar language likes other Southeast Asia languages that do not place spaces between words. Therefore, the system used Myanmar Word Segmenter (MWS) which is implemented in UCSYNLP Lab which is available for research purpose. Firstly, the system used translation probabilities without additional morphology analysis of Myanmar language. There are many unknown words in this process. Some of the unknown words occur due to the inflective nature of Myanmar Language. Languages may be divided into three broad categories: isolating, agglutinative and inflective languages. Isolating languages, such as Chinese, have little or no morphology and thus do not benefit from morphologically analysis. Agglutinative languages, also known as agglomerative or compounding languages, are those in which basic roots and words can be combined to make new words. These languages, such as Turkish or Finnish, tend to have many morphemes. Inflectional morphemes are used to modify a word to reflect information such as tense. Myanmar language may be agglutinative language and inflective language because Myanmar word can be combined to make new word. When a form of a word does not occur in the training 25th Pacific Asia Conference on Language, Information and Computation, pages 130–139 130 data, the system is unable to translate it. According to experimental result, the Out-of-Vocabulary (OOV) rate exceeds 50% for tested dataset with 2000 training sentences, which means that half of the words in test set are not present in the training set. Most of the OOV …

Content truncated to 500 words