Inferring Entity Types from Enumerative Descriptions

• Item type: SIG Technical Reports(1)
• 公開日: 2013-11-19
• タイトル: Inferring Entity Types from Enumerative Descriptions
• タイトル / 言語: en
• タイトル: Inferring Entity Types from Enumerative Descriptions
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: 情報抽出
• 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: 早稲田大学大学院情報生産システム研究科
• 著者所属: 早稲田大学大学院情報生産システム研究科
• 著者名: QianChen; 岩井原瑞穂
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Entity class matching has many real world applications, especially in entity clustering, de-duplication and efficient query processing. Current methods to extract entities from text usually disregard horizontal relationships, concentrating on either a prototype-based entity which lacks relationship between two clusters or a terminological entity where only hierarchical relationships are considered. We focus on enumerative descriptions that enlist entity names, together with parent types, often occurring in web documents as listings and tables. We consider discovering entities and relationships from two strongly related enumerative descriptions. We propose a RDF schema-based algorithm to capture a probabilistic RDF graph from enumerative descriptions by assigning candidate labels to each related token. Our algorithm is iterative: We infer candidate labels from matching sequential patterns and infer patterns that match well with current instances by updating confidence score on labeling of tokens.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Entity class matching has many real world applications, especially in entity clustering, de-duplication and efficient query processing. Current methods to extract entities from text usually disregard horizontal relationships, concentrating on either a prototype-based entity which lacks relationship between two clusters or a terminological entity where only hierarchical relationships are considered. We focus on enumerative descriptions that enlist entity names, together with parent types, often occurring in web documents as listings and tables. We consider discovering entities and relationships from two strongly related enumerative descriptions. We propose a RDF schema-based algorithm to capture a probabilistic RDF graph from enumerative descriptions by assigning candidate labels to each related token. Our algorithm is iterative: We infer candidate labels from matching sequential patterns and infer patterns that match well with current instances by updating confidence score on labeling of tokens.
• 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: AN10112482
• 書誌情報: 研究報告データベースシステム（DBS）; 巻 2013-DBS-158, 号 25, p. 1-7, 発行日 2013-11-19
• 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