2026-10-14T18:09:28Z https://www.scidb.cn/oai

10.57760/sciencedb.j00133.00714 2026-09-20T09:58:02Z

2026-09-20 BIOS-CN and DiseaseKG medical datasets doi:10.57760/sciencedb.j00133.00714 en This article selects BIOS Chinese related subsets and DiseasyKG as the data basis for medical knowledge graph retrieval and generation experiments. Among them, the BIOS data comes from the Biomedical Informatics Ontology System, which is a large-scale biomedical knowledge graph automatically constructed by machine learning methods. It is structured and organized for terms, concepts, and relationships in biomedical literature, including Chinese and English terms, concept nodes, and relationship triplets. It can provide basic support for medical entity standardization, cross language term alignment, and large-scale knowledge path retrieval. Public information shows that BIOS currently contains approximately 4.1 million concepts, 7.4 million bilingual terms, and 7.3 million relationship triplets, making it suitable for verifying the retrieval efficiency and path compression capability of models in large-scale medical knowledge resource environments. DiseasyKG is derived from the common disease information knowledge graph based on cnSchema published by OpenKG, mainly covering medical entity types such as diseases, drugs, food, examination items, departments, drugs on sale, symptoms, and treatment methods. It organizes common disease knowledge through relationships such as disease symptoms, disease examinations, disease drugs, and disease treatments. This dataset is relatively small in size, containing approximately 44656 entities and 312159 relationships, making it more suitable for testing the stability of the model in Chinese medical scenarios, low-frequency disease problems, and small-scale knowledge graph environments. The two types of datasets complement each other in terms of scale, language features, and knowledge coverage: BIOS focuses on organizing large-scale biomedical concepts and relationships, while DiseasyKG focuses on knowledge structures of common Chinese diseases. Therefore, the two together form the experimental basis for the retrieval performance, path compression, evidence tracing, and cross dataset stability verification in this article  BIOS; DiseaseKG; medical knowledge graph li jia long PUBLIC https://creativecommons.org/publicdomain/zero/1.0/ dataset Science Data Bank