Development and Evaluation of Kaldi Extension Tools with Python

• Item type: SIG Technical Reports(1)
• 公開日: 2019-11-29
• タイトル: Development and Evaluation of Kaldi Extension Tools with Python
• タイトル / 言語: en
• タイトル: Development and Evaluation of Kaldi Extension Tools with Python
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: ポスターセッション
• 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: Integrated Graduate School of Medicine, Engineering and Agricultural Sciences, University of Yamanashi
• 著者所属: Integrated Graduate School of Medicine, Engineering and Agricultural Sciences, University of Yamanashi
• 著者所属: Department of Industrial Information, Tsukuba University of Technology
• 著者所属: Graduate School of Systems and Informaation Engineering, University of Tsukuba
• 著者所属(英): en
• 著者所属(英): Integrated Graduate School of Medicine, Engineering and Agricultural Sciences, University of Yamanashi
• 著者所属(英): en
• 著者所属(英): Integrated Graduate School of Medicine, Engineering and Agricultural Sciences, University of Yamanashi
• 著者所属(英): en
• 著者所属(英): Department of Industrial Information, Tsukuba University of Technology
• 著者所属(英): en
• 著者所属(英): Graduate School of Systems and Informaation Engineering, University of Tsukuba
• 著者名: Yu, Wang; Hiromitsu, Nishizaki; Akio, Kobayashi; Takehito, Utsuro
• 著者名(英): Yu, Wang; Hiromitsu, Nishizaki; Akio, Kobayashi; Takehito, Utsuro
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: We have been developing extension tools of Kaldi, an automatic speech recognition toolkit, with Python language. A part of this toolkit works as wrapper of Kaldi, therefore, some operations, taking feature extraction and decoding with lattice as examples, are easily performed with Python code. In addition, our tools support training an acoustic model by using deep learning framework, such as Chainer. We evaluated our tools on TIMIT corpus and so on, and got a better ASR performance than other ASR systems in some ways.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: We have been developing extension tools of Kaldi, an automatic speech recognition toolkit, with Python language. A part of this toolkit works as wrapper of Kaldi, therefore, some operations, taking feature extraction and decoding with lattice as examples, are easily performed with Python code. In addition, our tools support training an acoustic model by using deep learning framework, such as Chainer. We evaluated our tools on TIMIT corpus and so on, and got a better ASR performance than other ASR systems in some ways.
• 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: 研究報告音声言語情報処理（SLP）; 巻 2019-SLP-130, 号 5, p. 1-5, 発行日 2019-11-29
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8663
• 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