| Presentation | 2018-12-10 Evaluation of Japanese end-to-end speech synthesis method inputting kana and prosodic symbols Kiyoshi Kurihara, Nobumasa Seiyama, Tadashi Kumano, Atsushi Imai, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | The purpose of this study was to conduct end-to-end text-to-speech synthesis in Japanese; we developed a system that uses Japanese kana characters and prosodic symbols as input for text-to-speech synthesis of the encoder-decoder model. Full-context phoneme labels are currently used for learning data in our system of statistical parametric speech synthesis, which uses a deep neural network (DNN) operating on broadcasting content. The data are composed of phonemes, phoneme boundary positions, accents, and accent phrases and are calculated automatically by executing language processing and forced alignment. The problem with this method is the high cost of manual correction for high-quality text-to-speech synthesis. In contrast, the proposed method does not prepare phoneme boundary positions. It was found that the workload of correcting the phoneme boundary position, which is particularly expensive, could be omitted. Japanese kana characters and prosodic symbols are simple description methods for representing speech. Hence, it is possible to directly compose kana and prosody and to arrange data without preparing dedicated software. In this study, we employed Japanese kana characters and prosodic symbols as the input method for text-to-speech synthesis by learning the encoder-decoder model. Subjective evaluation experiments indicated that Japanese text-to-speech results were equivalent in quality to those of synthesis using the acoustic features of the original speech. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Statistical parametric speech synthesis / End-to-end speech synthesis / Prosodic symbols / Encoder-Decoder model |
| Paper # | SP2018-49 |
| Date of Issue | 2018-12-03 (SP) |
| Conference Information | |
| Committee | NLC / IPSJ-NL / SP / IPSJ-SLP |
|---|---|
| Conference Date | 2018/12/10(3days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Waseda Univ. Nishiwaseda Campus |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | The 5th Natural Language Processing Symposium & The 20th Spoken Language Symposium |
| Chair | Takeshi Sakaki(Hottolink) / / Yoichi Yamashita(Ritsumeikan Univ.) |
| Vice Chair | Mitsuo Yoshida(Toyohashi Univ. of Tech.) / Kazutaka Shimada(Kyushu Inst. of Tech.) / / Akinobu Ri(Nagoya Inst. of Tech.) |
| Secretary | Mitsuo Yoshida(Ryukoku Univ.) / Kazutaka Shimada(NTT) / / Akinobu Ri(Kyoto Univ.) / (Meijo Univ.) |
| Assistant | Takeshi Kobayakawa(NHK) / Hiroki Sakaji(Univ. of Tokyo) / / Tomoki Koriyama(Tokyo Inst. of Tech.) / Satoshi Kobashikawa(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Natural Language Understanding and Models of Communication / Special Interest Group on Natural Language / 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 end-to-end speech synthesis method inputting kana and prosodic symbols |
| Sub Title (in English) | |
| Keyword(1) | Statistical parametric speech synthesis |
| Keyword(2) | End-to-end speech synthesis |
| Keyword(3) | Prosodic symbols |
| Keyword(4) | Encoder-Decoder model |
| 1st Author's Name | Kiyoshi Kurihara |
| 1st Author's Affiliation | Japan Broadcasting Corporation(NHK) |
| 2nd Author's Name | Nobumasa Seiyama |
| 2nd Author's Affiliation | Japan Broadcasting Corporation(NHK) |
| 3rd Author's Name | Tadashi Kumano |
| 3rd Author's Affiliation | Japan Broadcasting Corporation(NHK) |
| 4th Author's Name | Atsushi Imai |
| 4th Author's Affiliation | Japan Broadcasting Corporation(NHK) |
| Date | 2018-12-10 |
| Paper # | SP2018-49 |
| Volume (vol) | vol.118 |
| Number (no) | SP-354 |
| Page | pp.pp.89-94(SP), |
| #Pages | 6 |
| Date of Issue | 2018-12-03 (SP) |