Automatic Scoring Minimal-Pair Pronunciation Drills by Using Recognition Likelihood Scores and Phonological Features

Lei Chen, Qianyong Gao, Qiubing Liang, Jiahong Yuan, Yang Liu

In the mispronunciation detection task using automatic speech recognition (ASR) technology, a recent trend is utilizing phonological features (PFs). PFs have an advantage in generating pronunciation correction feedback comparing with the likelihood features from the ASR framework. However, previous studies only compared PFs with one type of likelihood feature, goodness of pronunciation (GOP). In this paper, we conducted a more thorough comparison by including more likelihood features proposed in the previous literature. Our experiments showed that PFs brought additional performance gains over basic likelihood features, but not for the feature set containing log likelihood ratio (LLR) features. Our findings are helpful to the community for a better understanding of the contributions of PFs to the mispronunciation detection task.

doi: 10.21437/SLaTE.2019-6

Cite as: Chen, L., Gao, Q., Liang, Q., Yuan, J., Liu, Y. (2019) Automatic Scoring Minimal-Pair Pronunciation Drills by Using Recognition Likelihood Scores and Phonological Features. Proc. 8th ISCA Workshop on Speech and Language Technology in Education (SLaTE 2019), 25-29, doi: 10.21437/SLaTE.2019-6

@inproceedings{chen19_slate,
  author={Lei Chen and Qianyong Gao and Qiubing Liang and Jiahong Yuan and Yang Liu},
  title={{Automatic Scoring Minimal-Pair Pronunciation Drills by Using Recognition Likelihood Scores and Phonological Features}},
  year=2019,
  booktitle={Proc. 8th ISCA Workshop on Speech and Language Technology in Education (SLaTE 2019)},
  pages={25--29},
  doi={10.21437/SLaTE.2019-6}
}