Presentation 1995/11/16
A study of language modeling for speech recognition using co-occurent word model
Mitsuru ENDO, Tatsuro ITO, Masakatsu HOSHIMI, Katsuyuki NIYADA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper we propose a new language model based on word pair co-occurrence, which is intended to realize high coverage, less over generation and easy task production. The co-occurrent word model approximate word N-gram probability with large N and it's constraint is verified to be stronger than 4-gram model's by experiments. furthermore, to achieve high coverage the Co-occurrent word model was modified to release word order and vocabulary constraints and to handle unknown words.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) speech recognition / language modeling / word n-gram model / co-occurrent word model
Paper # SP95-75
Date of Issue

Conference Information
Committee SP
Conference Date 1995/11/16(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 Speech (SP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A study of language modeling for speech recognition using co-occurent word model
Sub Title (in English)
Keyword(1) speech recognition
Keyword(2) language modeling
Keyword(3) word n-gram model
Keyword(4) co-occurrent word model
1st Author's Name Mitsuru ENDO
1st Author's Affiliation Matsushita Research Institute Tokyo, Inc.()
2nd Author's Name Tatsuro ITO
2nd Author's Affiliation Matsushita Research Institute Tokyo, Inc.
3rd Author's Name Masakatsu HOSHIMI
3rd Author's Affiliation Matsushita Research Institute Tokyo, Inc.
4th Author's Name Katsuyuki NIYADA
4th Author's Affiliation Matsushita Research Institute Tokyo, Inc.
Date 1995/11/16
Paper # SP95-75
Volume (vol) vol.95
Number (no) 355
Page pp.pp.-
#Pages 8
Date of Issue