SemTabNet

该数据集包含了超过10万个标注的表格，专为从ESG报告中提取定量事实及相关信息而设计。该数据集支持使用深度学习模型进行直接和间接陈述的提取，并包含了一种通过随机排列行和列来增强数据的方法。其规模宏大，包含超过10万个表格、84,890个独立单元格以及42,982条陈述。该数据集的任务是实现对表格中通用信息的提取。

This dataset contains over 100,000 annotated tables, specifically designed for extracting quantitative facts and relevant information from ESG reports. It supports the extraction of both direct and indirect statements via deep learning models, and incorporates a data augmentation method through random permutation of rows and columns. Boasting a substantial scale, it includes over 100,000 tables, 84,890 individual cells, and 42,982 statements. The core task of this dataset is to enable the extraction of general information from tables.