Presentation 2003/12/12
A Word-spotting Hypothesis Testing for Accepting/Rejecting Continuous Speech Recognition Output
Frank K. SOONG, Wai-Kit LO, Satoshi NAKAMURA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The word rejection problem in speech recognition is formulated in a framework of word-spotting, where a spotted word is verified through a binary, acceptance/rejection decision. A generalized word posterior probability (GWPP), used as the sole confidence measure, is computed in a word graph, via the forward-backward algorithm or in an N-best list, using string likelihoods. The GWPP is further enhanced by incorporating all spotted words with the same word ID and overlapped time registrations. When tested on the Japanese BTEC speech database, the confidence error rate is significantly reduced, from 23.76% to 17.78% and 20.18% to 15.57% for the two test data sets, respectively.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) confidence measure / word posterior probability / large vocabulary continuous speech recognition
Paper # NLC2003-98
Date of Issue

Conference Information
Committee NLC
Conference Date 2003/12/12(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Word-spotting Hypothesis Testing for Accepting/Rejecting Continuous Speech Recognition Output
Sub Title (in English)
Keyword(1) confidence measure
Keyword(2) word posterior probability
Keyword(3) large vocabulary continuous speech recognition
1st Author's Name Frank K. SOONG
1st Author's Affiliation Spoken Language Translation Research Laboratories, ATR()
2nd Author's Name Wai-Kit LO
2nd Author's Affiliation Spoken Language Translation Research Laboratories, ATR
3rd Author's Name Satoshi NAKAMURA
3rd Author's Affiliation Spoken Language Translation Research Laboratories, ATR
Date 2003/12/12
Paper # NLC2003-98
Volume (vol) vol.103
Number (no) 518
Page pp.pp.-
#Pages 6
Date of Issue