Project Objectives

This project aims to develop an Internet-based, computer-aided pronunciation training (CAPT) platform to help our students improve their English pronunciation. CAPT offers a round-the-clock, self-paced and personalized platform for the learner to exercise on pronunciation drills. We focus on language transfer effects due to Chinese (the mother tongue) in English (the target language) acquisition. We adapted automatic speech recognition (ASR) technologies to focus on the detection and diagnoses of the identified mispronunciations. ASR is integrated into an Internet-based, interactive CAPT system that can be accessed concurrently by a large number of learners for pronunciation drills. State-of-the-art ASR cannot guarantee perfect accuracies due to the intrinsic variability of input speech caused by the speaker, the environment and equipment. Nevertheless, automatic mispronunciation detection can still complement teacher-fronted instructions reasonably well.

Description of process, outcomes or deliverable

We have adapted automatic speech recognition technologies to enable automatic detection and diagnosis of mispronunciations. Our recognition framework can also perform probabilistic pronunciation scoring of the learner’s input speech. The technologies have been incorporated into an interactive system, named Enunciate, that is accessible on the Internet. We have designed lessons in the system to support self-directed pronunciation drills for the learner.

Evaluation

Obtained empirical evaluation results, through using online questionnaires, from learners interacting with the CAPT system both inside and outside of the classroom.

Dissemination, diffusion and impact

The system has been used in classes offered by the Faculty of Education and also the English Language Teaching Unit at CUHK. It is also accessible across campus, as well as the learning facility offered by the Independent Learning Centre. A demonstration of the system is also available on YouTube.