音声対話データの分析と発話理解への応用

• Item type: SIG Technical Reports(1)
• 公開日: 1994-10-21
• タイトル: 音声対話データの分析と発話理解への応用
• タイトル / 言語: en
• タイトル: Analysis and Application of Dialogue Corpus to Utterance Understanding
• 言語: jpn
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: 東京工業大学
• 著者所属: 東京工業大学
• 著者所属: 電子技術総合研究所
• 著者所属: 東京工業大学
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者所属(英): en
• 著者所属(英): Electrotechnical Laboratory
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者名: 上條俊一; 秋葉, 友良; 伊藤, 克亘; 田中, 穂積
• 著者名(英): Shunichi, Kamijo; Tomoyosi, Akiba; Katunobu, Itou; Hozumi, Tanaka
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: This paper describes the analysis of transcripted user's uttrances in the dialogues between human and machine obtained by Wizard of Oz simulation. The recorded data amounts to about 1 300 minutes. According to our analysis only 26 percent of user's utterances included filled pauses compared with the human-human dialogue in which about 50 percent of utterances included them and over 80 percent of filled pauses in human-machine dialogues were independent or found at the beginning of the sentences. Repairs were observed in 5 percent of the utterances and 73 percent of them were the repetitions or substitutions. We also report the analysis of other phenomena and suggest how to apply these results to utterance understanding.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: This paper describes the analysis of transcripted user's uttrances in the dialogues between human and machine obtained by Wizard of Oz simulation. The recorded data amounts to about 1,300 minutes. According to our analysis, only 26 percent of user's utterances included filled pauses compared with the human-human dialogue in which about 50 percent of utterances included them, and over 80 percent of filled pauses in human-machine dialogues were independent or found at the beginning of the sentences. Repairs were observed in 5 percent of the utterances and 73 percent of them were the repetitions or substitutions. We also report the analysis of other phenomena and suggest how to apply these results to utterance understanding.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10442647
• 書誌情報: 情報処理学会研究報告音声言語情報処理（SLP）; 巻 1994, 号 90(1994-SLP-003), p. 31-36, 発行日 1994-10-21
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会