@techreport{oai:ipsj.ixsq.nii.ac.jp:00216332, author = {鳥居, 克哉 and 中村, 覚 and 山田, 太造 and 稗方, 和夫 and Katsuya, Torii and Satoru, Nakamura and Taizo, Yamada and Kazuo, Hiekata}, issue = {8}, month = {Feb}, note = {本研究では,日本史学者の史料研究支援のために,史料群に対する可用性と有用性を高めるトピック抽出を自動で行うシステムの開発を行った.ルールベースにより抽出した人名及び N-gram や Sentencepiece によって分割した用語から Bag-of-Word を生成し,LDA (Latent Dirichlet Allocation) を適用することでトピック分析を行った.さらに,史料と人物索引表を入力としてこの一連の分析を行う Web システムをクラウド上に構築した.また,鎌倉時代の公卿である藤原(勘解由小路)経光が記した『民経記』を対象にこのシステムを利用し,トピック分析の結果が史実に整合していることが確認でき,有効性が示された., In this study, we developed a system that automatically extracts topics to increase the availability and usefulness of historical documents to support Japanese historians in their research on historical documents. We generated a Bag-of-Words from the names of people extracted by the rule base and the terms divided by N-gram and Sentencepiece., and applied LDA (Latent Dirichlet Allocation) to analyze the topics. In addition, we constructed a web system on the cloud to perform this series of analysis using historical documents and a person index table as input. In addition, we used this system to analyze the "Minkeiki" written by Fujiwara (Kadenokoji) Tsunemitsu, a kuge of the Kamakura period, and confirmed that the results of the topic analysis were consistent with the historical facts, demonstrating its effectiveness.}, title = {日本中世古記録を対象としたトピック抽出自動化システムの構築}, year = {2022} }