unicamp-dl/BR-TaxQA-R

该数据集旨在探索在巴西法律领域（特别是税法领域）中检索增强生成（RAG）系统的能力和性能。数据集基于巴西联邦税务局（RFB）自2016年以来每年发布的关于个人所得税的问答文档构建。数据集包括问题、答案和对法律文件的引用。还包括法律文件和CARF的案例法裁决。数据集以JSON文件格式组织，可以使用HuggingFace数据集库加载。

This dataset aims to explore the capabilities and performance of Retrieval Augmented Generation (RAG) systems in the Brazilian Legal Domain, particularly in Tax Law. The dataset is constructed from a Question & Answers document released annually by the Brazilian Federal Revenue Service (RFB) since 2016, focusing on Personal Income Tax. The dataset includes questions, answers, and references to legal documentation. Legal documents and case law rulings from CARF are also provided. The dataset is structured in JSON files and can be loaded using the HuggingFace datasets library.