Sci-Base
收藏资源简介:
Sci-Base是Sciverse科学数据基础的核心组成部分,专为科学人工智能(AI4S)社区设计,提供了一个大规模、高质量的科学知识基础数据集。该数据集通过深度解析和逻辑重构,将超过2500万份开放获取的科学论文和书籍转化为超过6000亿个纯令牌,覆盖数学与计算科学、物理学、化学、生命科学等10个核心科学领域。Sci-Base以其前所未有的规模、高精度解析(由MinerU技术支持)和丰富的科学实体而著称,知识截止日期至2026年3月,确保反映最新的科学突破。数据集结构清晰,每个实例代表一个科学文档,包含标题、作者、领域、发布日期等字段。使用Hugging Face的datasets库可轻松加载。数据集遵循CC-BY 4.0许可,用户需确保其应用符合原始材料的许可条款。
Sci-Base serves as the core component of the Sciverse Scientific Data Foundation, specifically designed for the scientific artificial intelligence (AI4S) community, and offers a large-scale, high-quality scientific knowledge base dataset. Through deep parsing and logical restructuring, it has converted over 25 million open-access scientific papers and books into over 600 billion pure Tokens, spanning 10 core scientific disciplines including mathematics and computational science, physics, chemistry, life sciences, and others. Sci-Base is renowned for its unprecedented scale, high-precision parsing powered by MinerU technology, and abundant scientific entities. Its knowledge cutoff date is March 2026, ensuring it captures the most recent scientific breakthroughs. The dataset has a well-defined structure: each instance represents a single scientific document, with fields including title, authors, discipline, publication date, and more. It can be readily loaded via Hugging Face's `datasets` library. The dataset is distributed under the CC-BY 4.0 license, and users are required to ensure that their usage of the dataset complies with the license terms of the original source materials.




