Using Phonological-Level Wav2Vec2 for Mandarin Automatic Mispronunciation Detection and Diagnosis

2026-06-20

AI Summary

This study addresses the limited diagnostic granularity of existing end-to-end approaches in Mandarin pronunciation error detection, which struggle to distinguish between segmental (initial/final) errors and tonal errors. To overcome this limitation, the authors propose a phonologically informed modeling framework based on Wav2Vec2-CTC that explicitly decomposes phonemes into segmental and tonal components and jointly models both aspects for fine-grained error detection. By incorporating a hierarchical phonological representation, the method substantially enhances diagnostic interpretability. Experimental results demonstrate that, compared to a baseline system using only undifferentiated phoneme labels, the proposed approach reduces the false acceptance rate by 10.1% and lowers the diagnostic error rate by 23.6%, highlighting its effectiveness in capturing nuanced pronunciation errors in Mandarin.

Abstract

Automatic mispronunciation detection and diagnosis (MDD) plays a crucial role in L2 Mandarin pronunciation learning. While end-to-end (E2E) based MDD methods have substantially improved phoneme-level detection accuracy, diagnostic feedback remains limited, as segmental and tonal errors are not explicitly separated. In this paper, we propose a phonological feature-based MDD framework that models both segmental and tonal attributes within a unified Wav2Vec2 CTC architecture. Experimental results show that the proposed method reduces the False Acceptance Rate (FAR) by 10.1% and the Diagnostic Error Rate (DER) by 23.6% compared with the phoneme-only baseline system. By decomposing phonemes into low-level phonological components, the proposed approach enables more detailed and interpretable diagnostic feedback for L2 learners.

Problem

• mispronunciation detection
• diagnostic feedback
• segmental errors
• tonal errors
• Mandarin pronunciation

Innovation

• phonological features
• Wav2Vec2
• mispronunciation diagnosis
• segmental-tonal modeling
• CTC architecture