A prototype of an adaptive Chinese pronunciation training system

Abstract

Many beginning-level Chinese learners, particularly those whose mother tongue is not a tone language find it challenging to learn Chinese pronunciations and tones. Yet pronunciation practice in formal language class settings is limited. Therefore, it has become necessary to develop computer-assisted and adaptive training systems to help learners practice Chinese pronunciation outside of class. This article reports on the development and verification of a prototypical adaptive Chinese pronunciation training system that specifically focuses on pronunciation errors related to aspiration, retroflexion, and tones. We describe how we built and trained the pronunciation error detection system, and how the system detected pronunciation errors and determined the timing of pronunciation exercises. The verification results of the system's performance demonstrated that it achieves a relatively high accuracy rate in detecting pronunciation errors and it has potential to help learners overcome specific weaknesses in Chinese pronunciation. We conclude with suggestions for further research on and development of an adaptive Chinese pronunciation training system.