ArchivalQA
收藏资源简介:
ArchivalQA是由京都大学创建的大型问答数据集,包含532,444个问题-答案对,专为历史新闻问答设计。数据集根据问题难度和时间表达式的包含情况分为四个子部分,有助于训练和测试不同强项和能力的开放域问答系统。该数据集利用自动问题生成和一系列精心设计的过滤步骤来生成高质量、无歧义的问题,适用于教育领域,如支持考试问题的生成。
ArchivalQA is a large-scale question answering dataset developed by Kyoto University, containing 532,444 question-answer pairs and specifically designed for historical news-based question answering. The dataset is divided into four subsets based on question difficulty and the presence of temporal expressions, which helps train and test open-domain question answering systems with varying strengths and capabilities. This dataset employs automatic question generation and a series of meticulously designed filtering steps to produce high-quality, unambiguous questions applicable to educational scenarios such as assisting in the generation of exam questions.




