"Minna de Honkoku" A Crowd-Sourced Transcription Project for Japanese Historical Materials

Abstract:

"Minna de Honkoku" (https://honkoku.org/) was launched as an online citizen science project to transcribe earthquake-related historical materials from Earthquake Research Institute Library, the University of Tokyo. In Japan, almost all the documents are written in Kuzushi-ji. Kujzushi-ji is writing style used before ~1900. Since the style is different from that of modern Japanese, transcription is necessary to use the historical documents as data for earthquake research. "Minna de Honkoku" consists of viewer of document image and vertical (Japanese-style) editor for transcription. Users can input transcribed texts viewing its image. The ranking of characters transcribed is displayed to keep motivation of users. The system adopted learning-based approach to encourage people to participate. Until July 2020, over 6,500 people registered for the project. We have loaded 1851 documents on the project. Total number of characters transcribed is about 10 million.

On July 2019, we have upgraded the system of Minna de Honkoku to support IIIF, International Image Interoperability Framework. We can now transcribe broader range of manuscripts on digital archives adopting IIIF. We have extended the subjects of the project, which now covers wide variety of historical materials as well as earthquake-related materials. We have also implemented AI-assisted transcription. Two AI engine suggest transcription of character, when users asked to help. The project help participants to obtain skills for Kuzishi-ji reading, knowledge on historical documents, historical disasters, idea to cooperate with AI and so on.