OUTPUT PROGRAM, INFORMATION PROCESSING DEVICE, AND OUTPUT CONTROL METHOD

• Main Authors: IDEUCHI MASAO, KOBAYASHI TAMANA, KATAOKA MASAHIRO
• Format: Patent
• Language: English; Japanese
• Published: 19.09.2019
• Subjects: ACOUSTICS; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

To improve the accuracy of speech recognition when speech recognizing a composition composed of a plurality of sentences.SOLUTION: An information processing device 100 accepts a phoneme string for a composition composed of a plurality of sentences, and specifies a sentence corresponding to a specific phoneme or phoneme string included in the accepted phoneme string. The information processing device 100 refers to sentence HMM data 143 in which co-occurrence information of sentence for a word is stored in association with the word, and specifies a word, among the words, for which the co-occurrence information of the specified sentence satisfies a criterion. The information processing device 100 changes a specific phoneme or phoneme string included in the accepted phoneme string to the specified word and thereby generates a composition that corresponds to the accepted phoneme string and outputs the generated composition.SELECTED DRAWING: Figure 2 【課題】複数の文で構成される文章を音声認識する場合に、音声認識の精度を向上する。【解決手段】情報処理装置１００は、複数の文を有する文章に対する音素列を受け付け、受け付けた音素列に含まれる特定の音素又は音素列に応じた文を特定する。情報処理装置１００は、単語に対する文の共起情報を該単語に対応付けて記憶する文ＨＭＭデータ１４３を参照して、単語のうち、特定した文の共起情報が基準を満たす単語を特定する。情報処理装置１００は、受け付けた音素列に含まれる特定の音素又は音素列を、特定した単語に変更することにより、受け付けた音素列に応じた文章を生成し、生成した文章を出力する。【選択図】図２