Efficient Phoneme Set Design Using Phonetic Decision Tree in Dialogue-Based English CALL Systems for Japanese Students

摘要

Taking consideration of the characteristics of users for dialogue-based CALL systems which aim at eliciting more speech productions in order to improve their speaking skill, we propose a reduced phoneme set consisting of more coarsely classified phonemes instead of the canonical phoneme set of a second language. This reduced phoneme set provides an inherently more appropriate model for error-prone pronunciations by second language speakers. In this study, we used a phonetic decision tree based top-down sequential splitting method to generate the coarse phoneme set. We then applied the proposed method to the task of recognizing English speech produced by Japanese students. Experimental results showed that a speech recognizer using the reduced phoneme set not only achieved higher recognition accuracy but also worked at higher speeds than that of the canonical phoneme set.