Presentation | 1998/12/11 Language Modeling and Acoustic Modeling for Automatic Transcription of Japanese Broadcast-News Speech Katsutoshi Ohtsuki, Sadaoki Furui, Naoyuki Sakurai, Atsushi Iwasaki, Zhi-Peng Zhang, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In this paper, we report on language modeling and acoustic modeling studies for broadcast-news speech recognition. We have been working on the development of a large-vocabulary continuous speech recognition(LVCSR)system for Japanese broadcast-news speech transcription. We constructed a language model that depended on the readings of words, whereas, usual language models depend on written words. In broadcast-news, each speaker utters several sentences in succussion, therefore we applied on-line speaker adaptation which is applied after identifying a speaker of the sentence. The reading-dependent language model reduced word error rate by about 10%, and the on-line speaker adaptation reduced word error rate by about 15%. We propose a new formulation for speech recognition, which maximizes the a posteriori probability of the speaker's intended message for a given observed acoustic sequence. We applied this formulation to rescoring N-best hypotheses and achieved better results with it. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | LVCSR / broadcast-news speech / n-gram / on-line speaker adaptation / message-driven speech recognition |
Paper # | NLC98-44,SP98-108 |
Date of Issue |
Conference Information | |
Committee | NLC |
---|---|
Conference Date | 1998/12/11(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 | Natural Language Understanding and Models of Communication (NLC) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Language Modeling and Acoustic Modeling for Automatic Transcription of Japanese Broadcast-News Speech |
Sub Title (in English) | |
Keyword(1) | LVCSR |
Keyword(2) | broadcast-news speech |
Keyword(3) | n-gram |
Keyword(4) | on-line speaker adaptation |
Keyword(5) | message-driven speech recognition |
1st Author's Name | Katsutoshi Ohtsuki |
1st Author's Affiliation | NTT Human Interface Laboratories() |
2nd Author's Name | Sadaoki Furui |
2nd Author's Affiliation | Tokyo Institute of Technology, Department of Computer Science |
3rd Author's Name | Naoyuki Sakurai |
3rd Author's Affiliation | Tokyo Institute of Technology, Department of Computer Science |
4th Author's Name | Atsushi Iwasaki |
4th Author's Affiliation | Tokyo Institute of Technology, Department of Computer Science |
5th Author's Name | Zhi-Peng Zhang |
5th Author's Affiliation | Tokyo Institute of Technology, Department of Computer Science |
Date | 1998/12/11 |
Paper # | NLC98-44,SP98-108 |
Volume (vol) | vol.98 |
Number (no) | 461 |
Page | pp.pp.- |
#Pages | 7 |
Date of Issue |