Low-cost and Automatic Linked Data Conversion for Machine-readable Bibliographies of Cultural Properties

2021 Fiscal Year Final Research Report

• Project/Area Number: 19K20630
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 90020:Library and information science, humanistic and social informatics-related
• Research Institution: Saga University
• Principal Investigator: Natsuko Yoshiga 佐賀大学, 地域学歴史文化研究センター, 研究機関研究員 (70457498)
• Project Period (FY): 2019-04-01 – 2022-03-31
• Keywords: 江戸期古記録 / シチズンサイエンス / 深層学習 / 固有表現抽出 / 単語分散表現 / 機械可読化

Outline of Final Research Achievements

In order to understand the contents of the many historical records that have survived in the local area, it is essential to have the expertise to read and understand the handwritten, kuzushiji documents in the literary language of the era. Currently, there are only a few people with such expertise, and this is a major barrier to analyzing local history and culture. In this study, we developed a method to mechanically extract unique expressions (keywords) such as names of people, events, and places from reprinted texts as much as possible, which can be the key to deciphering the text. In the first half of the study, we asked citizens who were originally interested in local materials to extract local-specific expressions that could not be found on the Internet from the perspective of citizen science. In the second half of the research, we established a method to extract unique expressions with high accuracy by using deep learning methods based on the manual extraction results.

Free Research Field

人文情報学

Academic Significance and Societal Importance of the Research Achievements

近年、我々の身の回りで起きる出来事をデータ化し、社会課題の解決に活かすデータ駆動型社会への移行が加速している。郷土の歴史資料においても同様に単なる画像への電子化に留まらず、テキスト化・機械可読化することが、人手に余る大量の資料の解析に必要であるとの認識が高まっている。本研究では、地域の歴史を知り守りたいと思う市民の助けと機械学習の力で歴史資料を低コストに機械可読化する手法を確立した。さらに、従来のデジタルアーカイブにおけるデータ提供者と利用者の役割を超えて関係者全員が文化財データを構築していく市民科学の実践にも貢献した。