Title:
テキスト補正装置、テキスト補正方法、およびプログラム
Document Type and Number:
Japanese Patent JP6552999
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To efficiently generate transcribed text that includes unnecessary words.SOLUTION: A morpheme analysis unit 1 analyzes the morpheme of transcribed text that was transcribed from a spoken voice. A grammar generation unit 2 inserts an unnecessary word in the morpheme analysis result for each morpheme boundary and generates a grammar model. A speech recognition unit 3 generates a plurality of speech recognition result candidates in which the speech data of spoken voices is speech recognized using the grammar model. A recognition result selection unit 4 calculates, for each speech recognition result candidate, the degree of similarity of an unnecessary word that is a syllable train included in the speech recognition result candidate to a word that follows the unnecessary word, and selects a speech recognition result of the spoken voice from the speech recognition result candidates on the basis of the calculated degree of similarity.SELECTED DRAWING: Figure 1
More Like This:
Inventors:
Takashi Nakamura
Hirokazu Masataki
Hirokazu Masataki
Application Number:
JP2016131807A
Publication Date:
July 31, 2019
Filing Date:
July 01, 2016
Export Citation:
Assignee:
Nippon Telegraph and Telephone Corporation
International Classes:
G10L15/06; G10L15/22
Domestic Patent References:
| JP2010243914A | ||||
| JP2007226117A |
Other References:
土屋雅稔他,フィラー予測モデルに基づくフィラー付き言語モデルの構築,第1回音声ドキュメント処理ワークショップ講演論文集,豊橋技術科学大学メディア科学リサーチセンター/情報処理学会音声言語情報処理研究会,2007年 2月,p.81-88
秋田祐哉他,統計的機械翻訳の枠組みに基づく言語モデルの話し言葉スタイルへの変換,電子情報通信学会技術研究報告,2005年12月,Vol.105,No.496,p.19-24
秋田祐哉他,統計的機械翻訳の枠組みに基づく言語モデルの話し言葉スタイルへの変換,電子情報通信学会技術研究報告,2005年12月,Vol.105,No.496,p.19-24
Attorney, Agent or Firm:
Naoki Nakao
Yukio Nakamura
Yoshimura Munehiro
Yukio Nakamura
Yoshimura Munehiro