BEE-spoke-data/bigpatent-all

BEE-spoke-data/bigpatent-all数据集是基于原始bigpatent数据集的一个子集，转换为HuggingFace格式，并对摘要列进行了重命名和清洗。该数据集包含三个配置：deduped、default和mini。deduped配置对摘要进行了去重，mini配置是deduped配置的一个约1GB版本，通过聚类和抽样方法得到。数据集适用于文本生成和摘要任务，语言为英文，包含法律相关的标签。

BEE-spoke-data/bigpatent-all dataset is a subset of the original bigpatent dataset, converted to the HuggingFace format with column renaming and cleaning applied to the summary column. It includes three configurations: deduped, default, and mini. The deduped configuration has the summary column deduplicated, and the mini configuration is a ~1GB version of the deduped dataset created using a clustering and sampling approach similar to minipile. The dataset is suitable for text generation and summarization tasks, is in English, and includes tags related to legal content.