Presentation | 2016-07-28 Evaluation of Japanese English DNN Acoustic Models with English Level Yuta Kawachi, Hirokazu Masataki, Taichi Asami, Yushi Aono, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In this paper, we propose an acoustic model that takes into consideration foreign language fluency level by extracting a representational vector for nativeness. Deep neural network acoustic models are commonlyused but their accuracy is not stable if they are trained using the non-native speech data. This is caused by the wide variation in non-native pronunciations, which are related to their foreign language fluency level. We tackled this problem by extending the acoustic feature vectors through the concatenation of the nativeness feature vectors. Nativeness is defined by bottleneck feature vectors of a deep neural network trained to distinguish languages. These additional feature vectors separate the input acoustic space of non-native speakers by their foreign language fluency levels to prevent confusion with features from different levels. We evaluated the proposed deep neural network acoustic model using Japanese English speech data, and show that the proposed model outperforms conventional DNN-HMM, which is trained using the same speech data. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | speech recognition / non-native / acoustic modeling / deep learning |
Paper # | SP2016-20 |
Date of Issue | 2016-07-21 (SP) |
Conference Information | |
Committee | SP / IPSJ-SLP |
---|---|
Conference Date | 2016/7/28(3days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Takinoyu Hotel |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Speech recognition and understanding, dialog system, etc. |
Chair | Kazunori Mano(Shibaura Inst. of Tech.) / Nobuaki Minematsu(Univ. of Tokyo) |
Vice Chair | Hiroki Mori(Utsunomiya Univ.) |
Secretary | Hiroki Mori(Kobe Univ.) / (Shizuoka Univ.) |
Assistant | Taichi Asami(NTT) / Kei Hashimoto(Nagoya Inst. of Tech.) |
Paper Information | |
Registration To | Technical Committee on Speech / Special Interest Group on Spoken Language Processing |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Evaluation of Japanese English DNN Acoustic Models with English Level |
Sub Title (in English) | |
Keyword(1) | speech recognition |
Keyword(2) | non-native |
Keyword(3) | acoustic modeling |
Keyword(4) | deep learning |
1st Author's Name | Yuta Kawachi |
1st Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
2nd Author's Name | Hirokazu Masataki |
2nd Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
3rd Author's Name | Taichi Asami |
3rd Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
4th Author's Name | Yushi Aono |
4th Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
Date | 2016-07-28 |
Paper # | SP2016-20 |
Volume (vol) | vol.116 |
Number (no) | SP-165 |
Page | pp.pp.1-6(SP), |
#Pages | 6 |
Date of Issue | 2016-07-21 (SP) |