Development of Data Mining Resources for Knowledge Discovery of Pre-Modern Japanese Concepts

点击翻译按钮获取中文摘要

We constructed the full text data of Chi part of Kojiruien (4,200 pages) attaching XML tags. The fourth proofed complete text data of volume I of Chi part is available with extended characters and the conversion table to Unicode. The volume II and III revised second and first, respectively ; besides, we digitized Kuzushi-ji yoorei jiten (Tokyodo Shoten), the standard dictionary for reading historical documents, to complement pre-modern terms.We developed a web-based system to browse Ten part text of Kojiruien and the full page images of whole Kojiruien (67,000 pages) with linking Kojiruien index (42,000 terms) and its phonetic index (64,000 entries). Through our research, electric information of basic pre-modern terms was developed. The Kojiruien index was converted into a thesaurus dictionary; we uploaded the thesaurus and a dictionary usable for Japanese morphological analysis to the website of our project. The web-based system of Kojiruien full text and page images are also opened at the International Research Center for Japanese Studies and the National Institute for Japanese Literature. In addition, we developed a Wiki-based prototype system and conducted a feasibility study of its effectiveness.As a case study of data mining for the Humanities, we conducted a data mining study to Renga Haikai database of the International Research Center for Japanese Studies. Through the investigation, we generated a collocation dictionary of pre-modern terms; by using the dictionary, we were successfully extracted a new finding on transition of hototogisu image in renga. Finally, we published the collocation dictionary of renga.