Thai phonetization of English words using English syllables

In Thai, it is not easy to find transcriptions of Roman characters such as English words. This leads to problems in Thai text-to-speech when facing with English words among Thai writing. Automatic English to Thai Grapheme-to-Phoneme conversion is proposed to tackle this problem. This paper presents an application of English syllable features to tone prediction model which is capable of classifying ambiguous tones and a refinement of post-processing module for phone prediction model. Experimental results show that the syllable and word accuracy are 76.03% and 53.93% respectively, which outperform all previous models.