Personality Recognition on Dyadic Interactions with Representation Learning

• Item type: SIG Technical Reports(1)
• 公開日: 2023-02-21
• タイトル: Personality Recognition on Dyadic Interactions with Representation Learning
• タイトル / 言語: en
• タイトル: Personality Recognition on Dyadic Interactions with Representation Learning
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: ショート・オーラル2
• 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: Tokyo Institute of Technology
• 著者所属: Yokohama City University
• 著者所属: 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): Yokohama City University
• 著者所属(英): en
• 著者所属(英): Tokyo Institute of Technology
• 著者名: Nathania, Nah; Takafumi, Koshinaka; Koichi, Shinoda
• 著者名(英): Nathania, Nah; Takafumi, Koshinaka; Koichi, Shinoda
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Personality computing explores methods of automatically measuring human traits to create a better understanding of the human psyche and thought processes. We examine conversations and interactions in dyadic environments through the perspective of representation learning to capture the psychological traits that compose a target's personality profile. We propose a bimodal speech-text model to predict scores for personality traits at a sentence level for the speakers using disentangled representations on speech and text. Our model outperforms current personality prediction methods using visual features and/or metadata on the UDIVA dataset's English subset.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Personality computing explores methods of automatically measuring human traits to create a better understanding of the human psyche and thought processes. We examine conversations and interactions in dyadic environments through the perspective of representation learning to capture the psychological traits that compose a target's personality profile. We propose a bimodal speech-text model to predict scores for personality traits at a sentence level for the speakers using disentangled representations on speech and text. Our model outperforms current personality prediction methods using visual features and/or metadata on the UDIVA dataset's English subset.
• 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 2023-SLP-146, 号 61, p. 1-6, 発行日 2023-02-21
• 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