Perceptual Evaluation of Pronunciation Quality for Computer Assisted Language Learning
暂无分享,去创建一个
In this paper, we propose a novel method of perceptual evaluation of pronunciation quality for Computer Assisted Language Learning used in e-learning. The overall score of the pronunciation quality is the combination of the matching score, the perceptual score and the asymmetric score. The matching score is the measure of the acoustic distortion of the test speech, the perceptual score models the perceived distortion by human in perception domain and the asymmetric score describes the asymmetric effect of the sensation of the deletion error and the insertion error in spoken English. The correlation coefficient between the predicted objective score and the subjective score by the experts is 0.75, which is advantageous over current methods based on HMM.
[1] Liu Runsheng. Phone-based pronunciation quality assessment algorithm , 2005 .
[2] Silke M. Witt,et al. Use of speech recognition in computer-assisted language learning , 2000 .
[3] Lou Boves,et al. Different aspects of expert pronunciation quality ratings and their relation to scores produced by speech recognition algorithms , 2000, Speech Commun..
[4] Vassilios Digalakis,et al. Combination of machine scores for automatic grading of pronunciation quality , 2000, Speech Commun..