Captioning Events in Tourist Spots by Neural Language Generation

• Item type: SIG Technical Reports(1)
• 公開日: 2019-06-06
• タイトル: Captioning Events in Tourist Spots by Neural Language Generation
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: 生成
• 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: 奈良先端科学技術大学院大学先端科学技術研究科
• 著者所属: 奈良先端科学技術大学院大学先端科学技術研究科／科学技術振興機構，さきがけ
• 著者所属: 奈良先端科学技術大学院大学先端科学技術研究科／現在，岐阜大学工学部所属
• 著者所属: 奈良先端科学技術大学院大学先端科学技術研究科
• 著者名: Mai, Nguyen; 吉野, 幸一郎; 鈴木, 優; 中村, 哲
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: We present an application that captions events in tourist attractions by summarizing various information sources in natural language descriptions. The system is divided into two parts: what-to-say, which summaries information into structured data, and how-to-say, which produces natural language captions from input meaning representation. In what-to-say, information from several information sources, such as infrared sensors and social media, are extracted into a semantic frame. In how-to-say, we utilized semantically-conditioned long short-term memory neural networks to generate natural language captions for giving information to users in an understandable way. An empirical evaluation of the system shows the quality of generated text across five automated metrics. The generated sentences are used in the application system for helping visitors. The subjective evaluation shows the usefulness of the proposed system.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: We present an application that captions events in tourist attractions by summarizing various information sources in natural language descriptions. The system is divided into two parts: what-to-say, which summaries information into structured data, and how-to-say, which produces natural language captions from input meaning representation. In what-to-say, information from several information sources, such as infrared sensors and social media, are extracted into a semantic frame. In how-to-say, we utilized semantically-conditioned long short-term memory neural networks to generate natural language captions for giving information to users in an understandable way. An empirical evaluation of the system shows the quality of generated text across five automated metrics. The generated sentences are used in the application system for helping visitors. The subjective evaluation shows the usefulness of the proposed system.
• 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: AN10115061
• 書誌情報: 研究報告自然言語処理（NL）; 巻 2019-NL-240, 号 10, p. 1-7, 発行日 2019-06-06
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8779
• 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