ChemDisGene

ChemDisGene数据集由马萨诸塞大学阿默斯特分校创建，包含80,000篇生物医学研究摘要，标注了化学物质、疾病和基因的提及，并远距离标注了18种类型的生物医学关系。数据集通过自动从CTD数据库中提取，旨在提高知识图谱的覆盖率，辅助生物医学研究，如疾病诊断、药物再利用等。创建过程中，使用了PubTator Central进行实体链接，提高了实体识别的准确性。该数据集适用于生物医学关系抽取模型的训练和评估，有助于解决生物医学文本中关系抽取的挑战。The ChemDisGene dataset was created by the University of Massachusetts Amherst. It contains 80,000 biomedical research abstracts annotated with mentions of chemicals, diseases, and genes, with distant annotation for 18 types of biomedical relationships. The dataset was automatically extracted from the Comparative Toxicogenomics Database (CTD) to improve the coverage of knowledge graphs and support biomedical research such as disease diagnosis, drug repurposing, and other related fields. During its development, PubTator Central was utilized for entity linking to enhance the accuracy of entity recognition. This dataset is suitable for training and evaluating biomedical relation extraction models, and helps address the challenges of relation extraction in biomedical texts.