ObliQA

该数据集名为ObliQA，包含了来自阿布扎比全球市场（ADGM）发布的监管文件中的27,869个问题。这一数据集旨在助力监管问答系统的开发。数据以JSON格式构建，并通过自然语言推理（NLI）进行验证，以确保问题与段落之间具有强烈的语义一致性。该数据集涵盖了约640,000个词汇，涉及40份监管文件，其中约21,000个问题与单一段落相关联，5,000个问题与多个段落相关联。其任务旨在进行监管合规性问题生成与段落检索。The dataset named ObliQA consists of 27,869 questions extracted from regulatory documents published by the Abu Dhabi Global Market (ADGM). This dataset is developed to support the advancement of regulatory question-answering systems. Structured in JSON format, the data is validated through Natural Language Inference (NLI) to guarantee robust semantic consistency between questions and their associated passages. The dataset contains approximately 640,000 tokens, covering 40 regulatory documents, with roughly 21,000 questions tied to a single passage and 5,000 questions linked to multiple passages. Its core tasks focus on regulatory compliance question generation and passage retrieval.

搜集汇总

背景与挑战

背景概述

ObliQA是一个基于义务的问答数据集，专为监管合规性研究设计，包含来自ADGM的监管文件，约640,000字，40份文件。数据集经过结构化处理，生成27,869个问题和答案对，适用于多种NLP应用。