TAT-QA
收藏资源简介:
TAT-QA是一个大规模的问答数据集,由新加坡国立大学等机构创建,专注于金融领域中结合表格和文本内容的混合数据。该数据集包含16,552个问题,涉及2,757个混合上下文,来源于182份真实的金融报告。数据集的创建过程涉及从报告中提取表格和相关文本,由具有金融知识的专业标注者生成问题和答案。TAT-QA特别强调数值推理,如加减乘除等,旨在推动混合数据问答技术的发展,特别是在需要复杂数值推理的场景中。
TAT-QA is a large-scale question answering dataset developed by institutions including the National University of Singapore, focusing on hybrid data combining tabular and textual content in the financial domain. This dataset contains 16,552 questions covering 2,757 hybrid contexts sourced from 182 real-world financial reports. The construction of TAT-QA involves extracting tables and associated texts from the reports, with professional annotators possessing financial domain knowledge generating the corresponding questions and answers. TAT-QA particularly emphasizes numerical reasoning tasks such as basic arithmetic operations including addition, subtraction, multiplication and division, aiming to advance the development of hybrid data question answering technologies, especially in scenarios requiring complex numerical reasoning.

- 1TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance新加坡国立大学 · 2021年



