This paper proposes a new workflow for structuring large-scale data, such as Pre-modern Japanese text “Bukan.” First, we define “Bukan” as a new type of historical sources called “time-series sources” that changes continuously over time, and interpret many versions associated with “Bukan” from the viewpoint of software engineering and make a mapping of those versions to the concepts of bibliography of Japanese old printed books. We then compare text-based and image-based approaches to the detection of difference, and propose a new concept “differential reading” that combines both the detection of difference, and differential transcription, to realize a workflow based on human-machine specialization, which is a key toward structuring “Bukan” The first preliminary result was released as “Bukan Complete Collection” on November 2017.

      @InProceedings{ jm17a,
      author = {Asanobu KITAMOTO and Hiroshi HORII and Misato HORII and Chikahiko SUZUKI and Kazuaki YAMAMOTO},
      title = {Structuring Time-Series Historical Sources by Human-Machine Specialization: Toward the Construction of Edo Information Platform Referring to “Bukan”},
      booktitle = {Proceedings of IPSJ SIG Computers and the Humanities Symposium 2017},
      pages = {273-280},
      year = 2017,
      month = 12,
      doi = {},
      note = { (in Japanese)},
      }