S2ORC

该数据集是一个经过精心筛选的、包含3100万篇英文科学论文的大型语料库。它覆盖了19个不同的科学领域，并用于创建一个异构文档图，以训练结构编码器模型。该数据集包含了3.54亿个文本段落，旨在应对科学领域中的文本段落检索与分类任务。

This dataset is a carefully curated large-scale corpus containing 31 million English scientific papers. It covers 19 distinct scientific domains and is utilized to construct heterogeneous document graphs for training structural encoder models. The dataset encompasses 354 million text paragraphs, and is designed to address text paragraph retrieval and classification tasks in the scientific domain.