Presentation | 2007/12/13 A Study on Normalization for Automatic Scoring of Pronunciation Proficiency of English Learners Kosuke Tsuji, Yoichi Yamashita, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This paper describes normalization methods for automatic scoring of pronunciation proficiency of English learners. Acoustic likelihoods calculated by speech recognition techniques are widely used for automatic scoring of pronunciation proficiency of language learners. It is well known that the normalization of acoustic likelihoods with posterior probabilities improves the accuracy of automatic scoring of pronunciation proficiency. The acoustic likelihood with an all-phone model which allows all transitions between phonemes approximate the probability of observed speech signals to calculate the posterior probability. This paper proposes acoustic likelihoods for each phoneme in stead of the average acoustic likelihood of the entire utterance using the all-phone model. Furthermore, an average of the acoustic likelihood is dependent on the phoneme even if a native speaker utters. This paper proposes the introduction of the average acoustic likelihood by native speakers to normalize pronunciation proficiency scores. The context dependency of the average acoustic likelihood by native speakers is also discussed. Evaluation experiments show that the proposed methods improve the correlation coefficient between estimated proficiency scores and human teachers' scores by about 0.1. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Posterior probability / Native speaker utterance / English / CALL / Pronunciation |
Paper # | NLC2007-59,SP2007-122 |
Date of Issue |
Conference Information | |
Committee | SP |
---|---|
Conference Date | 2007/12/13(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Speech (SP) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | A Study on Normalization for Automatic Scoring of Pronunciation Proficiency of English Learners |
Sub Title (in English) | |
Keyword(1) | Posterior probability |
Keyword(2) | Native speaker utterance |
Keyword(3) | English |
Keyword(4) | CALL |
Keyword(5) | Pronunciation |
1st Author's Name | Kosuke Tsuji |
1st Author's Affiliation | Graduate school of Science and Engineering, Ritsumeikan University() |
2nd Author's Name | Yoichi Yamashita |
2nd Author's Affiliation | College of Information Science and Engineering, Ritsumeikan University |
Date | 2007/12/13 |
Paper # | NLC2007-59,SP2007-122 |
Volume (vol) | vol.107 |
Number (no) | 406 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |