数字人文研究 ›› 2025, Vol. 5 ›› Issue (2): 88-109.

• 国际视野 • 上一篇    下一篇

“大家一起来翻刻” :基于协作学习的前近代日语史料众包转写平台

桥本雄太,日本国立历史民俗博物馆副教授;大邑润三,东京大学地震研究所助理教授;加纳靖之,东京大学地震研究所副教授;中西一郎,京都大学理学研究科教授;滨野未来,立命馆大学文学研究科日本史专业博士研究生;江晖(译者,通讯作者),中山大学外国语学院副教授。
  

  • 出版日期:2025-06-28 发布日期:2025-07-23

“Minna de Honkoku”: A Crowdsourced Transcription Platform for Premodern Japanese Materials Based on Collaborative Learning

  • Online:2025-06-28 Published:2025-07-23

摘要:

“大家一起来翻刻”是由日本京都大学团队开发、以日本前近代史料为对象的公众参与型转写平台。该平台于2017年以众包项目形式启动,最初专注于历史地震史料转写,2019年升级为通用日语史料转写平台。转写日本前近代史料需要解读“草书体”,为此“大家一起来翻刻”将参与者相互协作解读草体字的学习过程融入到平台设计当中,以此实现了大规模参与者的动员并维持其积极性以及参与者技能的提升。同时平台导入草书体自动识别AI辅助功能,提升复杂史料解读效率。平台现行版本已吸引约4100名参与者,完成超过2300份史料、共计4100万字的转写工作。这一模式不仅推动了历史文献的数字化保存与利用,还为公众参与学术研究提供了创新路径,具有重要的学术与社会价值。

关键词:

Abstract:

This paper introduces Minna de Honkoku, a citizen-participation-based transcription platform for premodern Japanese materials. Initially launched in 2017 as a collaborative transcription project for historical earthquake records, the platform was significantly updated in 2019 and expanded to cover a broader range of Japanese historical sources. Since the transcription of such materials requires deciphering kuzushiji (cursive script), Minna de Honkoku incorporates into its system design a learning process in which participants collaboratively acquire the skills needed for deciphering. This approach enables wide participation, sustains motivation, and enhances participants' skills. In addition, the system offers AI based support tools capable of automatically recognizing characters at the individual character or image level to assist with deciphering difficult kuzushiji. As of the current version, approximately 4,100 users have contributed to the transcription of over 2,300 items, totaling more than 41 million characters.

Key words:

中图分类号: