Presentation 1996/12/13
Speech understanding using a statistical translation language model
Tatsuo Matsuoka, Sadaoki Furui,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper describes a speech understanding method that uses a translation language model estimated automatically from a text corpus. The translation language model translates the natural language output by a speech recognition system into semantic language. For training the translation language model, words in natural and semantic languages are first clustered using a measure of word contextual similality. Natural and semantic languages are then modeled using grammar networks with the word clusters as nodes in the networks. Co-ocurrence probabilities of transitions in natural-language and semantic-language grammar networks are estimated as parameters of the translation language model. This method was shown to be very effective by experiments using the ARPA ATIS speech understanding evaluation task.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Speech understanding / Translation / Language modeling / Natural language / Semantic language
Paper # NLC96-52,SP96-83
Date of Issue

Conference Information
Committee NLC
Conference Date 1996/12/13(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Vice Chair

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) Speech understanding using a statistical translation language model
Sub Title (in English)
Keyword(1) Speech understanding
Keyword(2) Translation
Keyword(3) Language modeling
Keyword(4) Natural language
Keyword(5) Semantic language
1st Author's Name Tatsuo Matsuoka
1st Author's Affiliation NTT Human Interface Laboratories()
2nd Author's Name Sadaoki Furui
2nd Author's Affiliation NTT Human Interface Laboratories
Date 1996/12/13
Paper # NLC96-52,SP96-83
Volume (vol) vol.96
Number (no) 420
Page pp.pp.-
#Pages 8
Date of Issue