The Knowledge Graph Dataset for Martial Arts Culture

A high-quality martial arts knowledge dataset was constructed around the theme of martial arts for large-scale model applications, covering multiple core dimensions such as sects, characters, and encyclopedias. The dataset is mainly based on three platforms: the Chinese Martial Arts Association, the General Administration of Sport of China, and Baidu Baike. After manual screening, standardized processing, and quality monitoring, about 1000 raw data were generated, including martial arts encyclopedia knowledge, martial arts sects, martial arts masters and their relationships, martial arts rules and regulations, martial arts exchanges, and other aspects of information. Extracting over 25000 triplets of information from the constructed martial arts data using a large model, where the inheritor file includes the inheritor profiles and teacher-student relationships of each sect; The competition rules, regulations, policy and regulatory documents include the competition rules in the field of martial arts; The sect file contains information such as the introduction of each sect; The martial arts encyclopedia and Chinese martial arts encyclopedia documents contain knowledge about the meaning, development history, and significance of martial arts; The documents of martial arts masters and renowned martial arts schools include information on famous martial arts masters, schools, and introductions in the field of martial arts; The document on martial arts communication and information release contains detailed information on information exchange in the field of martial arts. This provides a solid data foundation for the application of question answering models in intelligent knowledge question answering tasks, and also promotes the digital dissemination of martial arts culture and the deep application of intelligent systems in traditional cultural fields.

本数据集围绕武术主题，面向大语言模型（Large Language Model，LLM）应用场景构建高质量武术知识数据集，覆盖门派、人物、百科等多个核心维度。本数据集主要依托三大平台构建：中国武术协会、国家体育总局及百度百科。经人工筛选、标准化处理与质量管控后，共生成约1000条原始数据，涵盖武术百科知识、武术门派、武术宗师及其师承关系、武术规章制度、武术交流等多类信息。并借助大模型从已构建的武术数据中提取超25000条信息三元组，其中：传承人档案涵盖各门派的传承人简介与师承关系；赛事规则、规章制度与政策文件板块包含武术领域的竞赛规程；门派档案板块涵盖各门派的基本介绍等信息；武术百科及中华武术百科文档板块收录武术的概念内涵、发展历程与价值意义等相关知识；武术宗师与知名武学流派文档板块收录武术领域知名宗师、流派及其相关介绍信息；武术传播与资讯发布文档板块则涵盖武术领域资讯交流的详细内容。本数据集可为问答模型应用于智能知识问答任务提供坚实的数据基础，同时助力武术文化的数字化传播，以及智能系统在传统文化领域的深度落地应用。