rootsautomation/ENTRANT

ENTRANT是一个HuggingFace镜像数据集，源自Zenodo记录（记录号10667088，许可证为CC-BY-4.0）。该数据集包含从约330,000份美国证券交易委员会（SEC）EDGAR文件中提取的670万结构化财务表格，涵盖了10种不同的文件类型。这些文件类型包括年度报告（10-K）、季度报告（10-Q）、外国年度报告（20-F）、IPO注册声明（S-1）、并购注册声明（S-4）、当前报告（8-K）、共同基金招股说明书（497）、基金注册后修正案（485BPOS）、过渡年度报告（10-KT）和外国政府年度报告（18-K）。数据集主要用于表格问答和文本分类任务，提供了表格的HTML渲染、单元格结构（如行、列、跨行跨列信息）以及元数据（如表格ID、文件类型、CIK、年份等）。原始论文发表于《Scientific Data》2024年，旨在支持财务表格的理解和分析。

ENTRANT is a HuggingFace mirror of the ENTRANT dataset (Zenodo record 10667088, CC-BY-4.0): 6.7 million structured financial tables extracted from ~330,000 SEC EDGAR filings spanning 10 filing types. The filing types include annual reports (10-K), quarterly reports (10-Q), foreign annual reports (20-F), IPO registration statements (S-1), merger/acquisition registrations (S-4), current reports (8-K), mutual fund prospectuses (497), fund registration post-effective amendments (485BPOS), transition annual reports (10-KT), and foreign government annual reports (18-K). The dataset is designed for table-question-answering and text-classification tasks, providing HTML-rendered tables, cell structures (with row, column, rowspan, colspan, text, and header/attribute flags), and metadata (such as table ID, filing type, CIK, year). The original paper was published in Scientific Data in 2024, focusing on table understanding in financial contexts.