Using Chi-Square Automatic Interaction Detector to Solve the Polysemy Problems in a Chinese to Taiwanese TTS System

We use chi-square automatic interaction detector (CHAID) to solve the polysemy problems in a Chinese to Taiwanese TTS system in this paper. Polysemy means there are words with more than one meaning or pronunciation, such as "???"(we), "???"(no), "???"(you), "???"(I), "???"(want) , and so on. The correct pronunciation of a word affects the comprehensibility (or clarity) and fluency of Taiwanese speech. We have shown that using language models to solve the polysemy problems with outstanding results. We use CHAID which has been found to be an elective approach for obtaining meaningful segments that are predictive of a K-category (nominal or ordinal) criterion variable to the polysemy problem in this paper. There are three words with polysemy problem be solved in this paper. They are "???"(no), "???"(up), and "???"(down). The numbers of kinds of pronunciation of the three words are 6, 3, and 4, respectively. Results show that the CHAID approach is over language models and decision list classifier.