Grammatical-based Word Segmentation Algorithm for Building Language Model in Myanmar ASR

In this paper a grammatical-based wordsegmentation algorithm is presented. As thenature of Myanmar language, wordsegmentation process is necessary for machinetranslation, speech recognition, speechsynthesis, etc. Many segmentation methods forMyanmar have been proposed for many reasons.Language model is the main part of computingon statistical speech recognition method. Thisword segmentation algorithm is proposed to usein building word-based language model forcontinuous speech recognition. We have used16950 sentences for word segmentation. In thisalgorithm, we used post-positional marker andconjunction as word token markers. The outputword lists are used to compute the probability ofword order sequence for speech recognitionsystem