Presentation 2009/12/14
Active learning using multiple recognizers for speech recognition
Yuzo Hamanaka, Tadashi Emori, Takafumi Koshinaka, Koichi Shinoda, Sadaoki Furui,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose an active learning method with multiple recognizers for large vocabulary continuous speech recognition. In this approach, the recognition results obtained from recognizers are used for selecting utterances. Here, a progressive search method is used for aligning sentences, and voting entropy is used as a measure for selecting utterances. Our method was evaluated by using 190-hour speech data in the Corpus of Spontaneous Japanese. It proved to be significantly better than random selection. It only required 60 h of data to achieve a word accuracy of 74%, while standard training(i.e., random selection)required 97 h of data. The recognition accuracy of our proposed method was also better than that of the conventional uncertainty sampling method using word posterior probabilities as the confidence measures for selecting sentences.
Keyword(in Japanese) (See Japanese page)
Keyword(in English)
Paper #
Date of Issue

Conference Information
Committee NLC
Conference Date 2009/12/14(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) Active learning using multiple recognizers for speech recognition
Sub Title (in English)
Keyword(1)
1st Author's Name Yuzo Hamanaka
1st Author's Affiliation Tokyo Institute of Technology()
2nd Author's Name Tadashi Emori
2nd Author's Affiliation NEC Informatec Systems, Ltd.
3rd Author's Name Takafumi Koshinaka
3rd Author's Affiliation Tokyo Institute of Technology:NEC Corporation
4th Author's Name Koichi Shinoda
4th Author's Affiliation Tokyo Institute of Technology
5th Author's Name Sadaoki Furui
5th Author's Affiliation Tokyo Institute of Technology
Date 2009/12/14
Paper #
Volume (vol) vol.109
Number (no) 355
Page pp.pp.-
#Pages 5
Date of Issue