Improvement of content missing documents using description required item identification technology

2018 Fiscal Year Final Research Report

• Project/Area Number: 26330252
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: Tottori University
• Principal Investigator: MURATA Masaki 鳥取大学, 工学研究科, 教授 (50358884)
• Project Period (FY): 2014-04-01 – 2019-03-31
• Keywords: 内容欠落文書の改善 / 文書推敲 / 記載必要項目 / 情報抽出 / 文生成 / 機械学習

Outline of Final Research Achievements

Rule-based techniques using items "comparison", "problem", "purpose" and "example" in article data could detect the absence of description with an F measure of 0.6 to 0.8. The technology was generalized. We made the technology to discover the presence or absence of the description of the important matters by using the frequent occurrence items in the documents similar to the documents to be modified as the important matters. Based on word level information extraction, we could detect absence of the description at F measure (performance) of about 0.8. We built a technology to obtain information related to the missing part from the web and present it to the user. We could build a technology to extract information based on sentences. In the newspaper and product information documents, we were able to carry out an experiment to find absence of the description at the sentence level. We found absence of the description in the F measure of about 0.5 to 0.8.

Free Research Field

自然言語処理

Academic Significance and Societal Importance of the Research Achievements

従来の推敲システムであまり扱われていなかった内容欠落文の修正の問題を扱ったため、推敲システムの扱える範囲が増え、推敲システムの発展に寄与する。さらに社会的に広く見れば、本技術が会話・通信・会議にも応用されることにより、種々のコミュニケーションにおける内容欠落文の減少による情報伝達率の上昇により、人間活動の大幅な効率化につながると期待される。