Development of a Japanese Pronunciation Training System for Learners ’ Speech Recognition

This study aims at developing a Japanese pronunciation training system, with special consideration to the learners ’ speech with errors. To give appropriate instructions for non-native speakers of Japanese, we, firstly analyzed the collected learners ’ speech and labeled all the errors into 39 categories. Secondly, to develop an original speech recognition system for learners ’ speech which contains various kinds of errors, we constructed a phoneme model based on both native and learners database. Using this model, we developed a Japanese pronunciation training system which can recognize and evaluate the speech of different L1 learners. We tentatively evaluated the system with Vietnamese learners to ensure the effectiveness and usability of the system