A Chinese Knowledge Graph Dataset in the Field of Scientific Fitness.

Abstract

To promote the development of scientific fitness research and practice, we propose the Chinese Knowledge Graph Dataset in the Field of Scientific Fitness (FitKG-CN). This knowledge graph contains over 10,000 fitness-related terms, categorized into eight main groups: body parts, items of exercise, fitness movement, equipment and tools, exercise goals, anatomical structures, nutrients, and technical terms. The construction of FitKG-CN is based on authoritative data sources, undergoing rigorous preprocessing, including noise removal, format standardization, and normalization of entities and relationships. The data is manually annotated on a professional platform and ultimately stored in a Neo4j graph database for visualization. Additionally, we trained a Chinese SpERT model using the manually annotated data to enhance the automation of data processing. The experimental results show that the model achieved an F1 score of 94.05% in entity recognition tasks and 82.00% in relation extraction tasks, validating the effectiveness of the model and improving the scalability of the dataset.