Section Article

The Bangla TTS System is constructed using Grapheme-Phoneme Rules and a Phonetic Dictionary.
Author(s): Rohit Bhalla

Abstract
The first phase in creating a top-notch Text to Speech synthesiser involves transforming text into its accurate phonetic representation. One possible strategy is to ascertain all the laws governing the conversion of graphemes to phonemes while another option is to compile a thorough dictionary including words in their corresponding phonetic forms. Both techniques are necessary but none of them is viable in isolation. To construct a complete dictionary it is important to examine extensive text corpora transcribe all words into their phonetic form and then include them into the dictionary. That is a colossal endeavour. Nevertheless there are always instances that deviate from every principle making it unfeasible to draft such legislation exhaustively in Bangla and other languages. To optimise the benefits of text-to-phonemic conversion it is advisable to use both of these methodologies in conjunction. This research elucidates the use of a hybrid approach in the Bangla Text to Speech Synthesis System. Here a voice synthesiser that operates based on formants as described earlier in this article is used as the central unit.