bilawalriaz/MedAtoms

MedAtoms是一个开放的原子医学事实数据集，专为构建可靠的医疗AI系统而设计。该子集包含从精选医学文献中提取的高置信度原子事实，每个事实都是一个独立、可验证的医学知识片段。主要特点包括：高置信度、原子粒度、医学专注和生产就绪。数据集包含465,643个事实，分为训练集（80%）、验证集（10%）和测试集（10%），涵盖34个医学领域和10,277个独特主题标签。每个事实条目包含唯一ID、文本、字数、字符数、医学领域、标签、核心标志和来源类型等信息。数据集适用于医学问答系统、医疗聊天机器人知识库、医学事实验证、医学NLP模型微调、检索增强生成（RAG）和医学教育工具等用途。

MedAtoms is an open dataset of atomic medical facts designed for building reliable healthcare AI systems. This subset contains high-confidence atomic facts extracted from curated medical literature. Each fact is a single, self-contained piece of verifiable medical knowledge. Key features include: high confidence only, atomic granularity, medical focus, and production ready. The dataset consists of 465,643 facts, split into training (80%), validation (10%), and test (10%) sets, covering 34 medical areas and 10,277 unique topic tags. Each fact entry includes a unique ID, text, word count, character count, medical area, tags, core flag, and source type. The dataset is intended for uses such as medical question answering systems, healthcare chatbot knowledge bases, medical fact verification, fine-tuning medical NLP models, Retrieval-Augmented Generation (RAG), and medical education tools.