IJSRP, Volume 2, Issue 9, September 2012 Edition [ISSN 2250-3153]
Chandan Sarma, Prof. P.H Talukdar
It is not possible to include all the words in a natural language for general text-to-speech system. Grapheme-to-phoneme conversion system is essential to pronounce a word which is out of vocabulary. Grapheme-to-phoneme rules play a vital role where lexical lookup fails. Though basic Grapheme-to-phoneme rules system is very simple yet it is very powerful for naturalness of a TTS system. Letter-to-sound rules may be hand written or maybe automatic depending on the language. We have worked on Bodo language. After a systematic study of Boro language we found that there is a systematic relationship between the written form of a Bodo word and its pronunciation. So, it is fairly easy to build letter-to-sound rules by hand for Bodo language. We have used a Bodo corpora of 5000 words and built letter-to-sound rules. These rules have been tested using Festival, a most popular speech synthesizer and applying these rules, we were able to produce correct pronunciations for approximately 89% of the words. Again, dialect variation also influences grapheme-to-phoneme conversion rules. This paper gives overview of Boro dialect variation and grapheme-to-phoneme conversion rules developed for Boro TTS system.