[目的/意义] 基于大语言模型，探索历史学领域循证知识库的构建与应用方法，旨在为历史学研究提供更加智能化的工具，促进数字人文时代的研究方法革新。[方法/过程] 首先，设计面向历史学的循证本体，建立证据语义组织框架；通过采用多阶段知识抽取策略，从历史文献中提取结构化的观点与证据。然后，结合向量检索与语言模型，进一步构建观点节点间的语义关联，形成历史学循证知识库。最后，采用多阶证据检索策略，逐层递进扩展证据检索范围，挖掘潜在的历史证据关联，支持历史学观点的智能化论证。[结果/结论] 基于历史学循证知识库构建的原型系统，在发掘深层历史证据和多维度论证方面展现了良好的效果，专家评测结果表明其具有较高的相关性、可靠性和实用性，展现出为历史研究提供智能化辅助的应用潜力。
[Purpose/Significance] Based on large language models, exploring the construction and application methods of an evidence-based knowledge base in the field of history can provide more intelligent tools for history research and promote the innovation of research methods in the digital humanities era. [Method/Process] First, an evidence-based history ontology was designed, establishing a framework for semantic organization of evidence. A multi-stage knowledge extraction strategy was employed to extract structured viewpoints and evidence from historical documents. Then, by combining vector retrieval with language models, semantic associations between viewpoint nodes were further constructed, forming an evidence-based knowledge base for history. Finally, a multi-level evidence retrieval strategy was adopted to progressively expand the scope of evidence retrieval, uncovering potential historical evidence associations to support the intelligent argumentation of historical viewpoints. [Result/Conclusion] The prototype system built on the evidence-based historical knowledge base has proven effective in uncovering deep historical evidence and enabling multidimensional argumentation. Expert evaluations suggest that the system demonstrates high relevance, reliability, and practicality, highlighting its potential to serve as an intelligent tool for historical research.
1] 杨克虎. 循证社会科学的产生、发展与未来[J].图书与情报, 2018(3): 1-10. (YANG K H. Evidence-based social science: the origin, development, and prospects[J]. Library & information, 2018(3): 1-10.)
[2] 赵坤, 郭君钰, 杨光, 等. Campbell图书馆简介[J].中国循证医学杂志, 2015, 15(1): 120-124. (ZHAO K, GUO J Y, YANG G, et al. Brief introduction of the Campbell Library[J]. Chinese journal of evidence-based medicine, 2015, 15(1): 120-124.)
[3] TUCKER A. A companion to the philosophy of history and historiography[M]. Hoboken: Wiley-Blackwell, 2009: 383-391.
[4] COLLINGWOOD R G. The idea of history[M]. Oxford: Oxford University Press, 1993: 1-21.
[5] 孙闻博. “二重证据法”与王国维的史学实践[J].中国史研究动态, 2022(2): 63-69. (SUN W B. Double evidence method and historiographical practice of Wang Guowei[J]. Trends of recent researches on the history of China, 2022(2): 63-69.)
[6] 杨骊, 叶舒宪. 四重证据法研究[M]. 上海: 复旦大学出版社, 2019: 18-19. (YANG L, YE S X. Quadruple-evidence method[M]. Shanghai: Fudan University Press, 2019: 18-19.)
[7] HARVARD UNIVERSITY, ACADEMIA SINICA, PEKING UNIVERSITY. China biographical database (CBDB)[DB/OL]. (2018-01-01)[2025-04-03]. https://projects.iq.harvard.edu/cbdb.
[8] BOL P K. The China historical geographic information system (CHGIS): choices faced, lessons learned[C]//Conference on historical maps and GIS. Nagoya: Nagoya University, 2007: 23-24.
[9] 夏翠娟, 刘炜, 陈涛, 等. 家谱关联数据服务平台的开发实践[J].中国图书馆学报, 2016, 42(3): 27-38. (XIA C J, LIU W, CHEN T, et al. A genealogy data service platform implemented with linked data technology[J]. Journal of library science in China, 2016, 42(3): 27-38.)
[10] 刘炜, 林海青, 夏翠娟. 数字人文研究的图书馆学方法:书目控制与文献循证[J].大学图书馆学报, 2018, 36(5): 116-123. (LIU W, LIN H Q, XIA C J. Bibliographic approach to digital humanities: Authority control and evidence-based practices[J]. Journal of academic library, 2018, 36(5): 116-123.)
[11] NILSEN P, SUNDEMO D, HEINTZ F, et al. Towards evidence-based practice 2.0: leveraging artificial intelligence in healthcare[J]. Frontiers in health services, 2024, 4: 1368030.
[12] 南京农业大学. 荀子系列大语言模型[EB/OL]. [2025-04-09]. https://github.com/xunzi-llm-of-chinese-classics/xunziallm. (Nanjing Agricultural University. Xunzi series large language model [EB/OL]. [2025-04-09]. https://github.com/xunzi-llm-of-chinese-classics/xunziallm.)
[13] 段丹洁. 古汉语大语言模型“AI太炎2.0”正式发布[EB/OL]. (2024-09-10)[2025-04-09]. https://www.sinoss.net/c/2024-09-10/642616.shtml. (DUAN D J. The classical Chinese large language model "AI Taiyan 2.0" officially released [EB/OL]. (2024-09-10) [2025-04-09]. https://www.sinoss.net/c/2024-09-10/642616.shtml.)
[14] NEO4J. Vector Index and Search[EB/OL]. [2025-05-01]. https://neo4j.com/labs/genai-ecosystem/vector-search/.
