High-Precision Search via Question Abstraction for Japanese Question Answering

スポンサーリンク

概要

• 論文の詳細を見る

• This paper explores the use of Question Abstraction, i.e., Named Entity Recognition for questions input by the user, for reranking retrieved documents to enhance retrieval precision for Japanese Question Answering (QA). Question Abstraction may help improve precision because (a) As named entities are often phrases, it may have effects that are similar to phrasal or proximity search; (b) As named entity recognition is context-sensitive, the named entity tags may help disambiguate ambiguous terms and phrases. Our experiments using several Japanese "exact answer" QA test collections show that this approach significantly improves IR precision, but that this improvement is not necessarily carried over to the overall QA performance. Additionally, we conduct preliminary experiments on the use of Question Abstraction for Pseudo-Relevance Feedback using Japanese IR test collections, and find positive (though not statistically significant) effects. Thus the Question Abstraction approach probably deserves further investigations.

• 一般社団法人情報処理学会の論文
• 2004-09-16

著者

• Koyama Makoto

  Knowledge Media Laboratory Toshiba Corporate R&d Center

• Saito Yoshimi

  Knowledge Media Laboratory Toshiba Corporate R&d Center

• SAKAI Tetsuya

  Knowledge Media Laboratory, Toshiba Corporate R&D Center

• KOKUBU Tomoharu

  Knowledge Media Laboratory, Toshiba Corporate R&D Center

• MANABE Toshihiko

  Knowledge Media Laboratory, Toshiba Corporate R&D Center

• Manabe Toshihiko

  Knowledge Media Laboratory Toshiba Corporate R&d Center

• Sakai Tetsuya

  Knowledge Media Laboratory Toshiba Corporate R&d Center

• Kokubu Tomoharu

  Knowledge Media Laboratory Toshiba Corporate R&d Center

• Sakai Tetsuya

  Knowledge Media Laboratory, Toshiba Corporate R&D Center

関連論文

• High-Precision Search via Question Abstraction for Japanese Question Answering
• High-Precision Search via Question Abstraction for Japanese Question Answering
• A Note on the Reliability of Japanese Question Answering Evaluation
• A Further Note on Evaluation Metrics for the Task of Finding One Highly Relevant Document(情報検索・分類,テーマ : 「デジタルアーカイブの活用(応用)」および一般)
• A Further Note on Evaluation Metrics for the Task of Finding One Highly Relevant Document(情報検索・分類,テーマ : 「デジタルアーカイブの活用(応用)」および一般)
• Controlling the Penalty on Late Arrival of Relevant Documents in Information Retrieval Evaluation with Graded Relevance

スポンサーリンク