[M-IS17] History X Earth and Planetary Science

[MIS17-P08] Consideration of Metrical Research about History of Astronomy Using Digital Archives: Survey of comet descriptions

*Harufumi Tamazawa^1,2 (1.Institute of Industrial Science, the University of Tokyo, 2.Kyoto City University of Arts)

Keywords:Degital Archive

Various natural phenomena are recorded in historical documents, which have been used as a type of data to provide information for understanding rare phenomena and long-term changes. In the field of astronomy, various astronomical phenomena have been picked up from historical documents and cataloged. As far as early modern Japanese historical materials are concerned, there are works such as Osaki's ``Early Modern Japanese Astronomical Materials,'' but the amount of remaining early modern materials means that the range that can be searched manually is limited. In recent years, OCR technology has developed, making it possible to perform mechanical surveys to a certain extent, and it is possible to conduct larger-scale surveys of phenomena. What level of response is currently being taken? Specifically, we will use approximately 80,000 classical materials, which are the results of the National Diet Library NDL Lab's FY2020 OCR text conversion experiment for classical materials, to search for specific astronomical terms, and to search for specific astronomical terms. Comparative studies can be made by checking records of known phenomena. The search results simply yielded 57 results for "suisei" (using old-type character) and 13 for "houkiboshi", but this clearly deviates from the number of known records, and the characteristics of the collected materials may also be considered. NDL OCR ver.3 was released in February 2023, and the results or resurvey were 60 and 14, respectively, indicating that improvements can be made by improving OCR accuracy and teacher models. On the other hand, it is also possible to conduct a survey of phenomena from a specific period based on previous research, and if you search for comets in 1823, you will find that they are recorded in the ``Kamosha Records''. On the other hand, there is also a record of a comet in 1843, but although it is recorded as a comet in writing, it is judged as a "calligraphy" star by OCR, and it is determined by the amount of learning in the OCR training data. There is also a possibility that it has become too lenient. It was confirmed that some degree of manual correction was necessary, and that quantitative research on the history of astronomy was now possible by improving the accuracy of previous research.