A part of the TTS-front end framework which serves to predict the correct pronunciation for the input polyphone characters.