Improving End-To-End Modeling for Mispronunciation Detection with Effective Augmentation Mechanisms

📖 Description

Impro ving End-To-End Modeling for Mispronunciation Detection with Effective Augmentation Mechanisms Tien-Hong Lo, Yao -Ting Sung and Berlin Chen National Taiwan Normal University, Taipei City, Taiwan {teinhonglo, sungtc, berlin} @ntnu.edu.tw Abstract ? Recently, end -to-end (E2E) model s, which allow to take spectral vector sequences of L2 (second -language) learners? utterances as input and produce the corresponding phone -level sequences as output , have attracted much research attention in developing mispronunciation detection (MD) system s. However, due to the lack of sufficient labeled speech data of L2 speakers for model estimation , E2E MD model s are prone to overfitting in relation to conventional ones that are built on DNN -HMM acoustic model s. To alleviate this critical issue , we in this paper propose two modeling strategies to enhance the discrimination capability of E2E MD model s, each of which can implicitly leverage the phoneti...

📊 Game Impacts

💭 Reactions

🔗 Sources

• [1] https://arxiv.org/abs/2110.08731