Computer-Assisted Pronunciation Training System for Atayal, an Indigenous Language in Taiwan

This research presents a computer-assisted pronunciation training system for Atayal, one of Taiwan's Indigenous languages. The core challenge addressed is modeling a low-resourced language and designing an automatic pronunciation training system for it. We employed a cross-lingual phoneme recognition model to assess learners' phonetic errors and used a comparison-based approach to evaluate learners' intonation quality. Subsequently, our system can visualize the position of pronunciation mistakes and provide feedback on mistake properties. After system testing, we observed that all the learners felt the system enhanced their learning experience.