MCScript

MCScript是由德国萨尔兰大学创建的一个大型数据集，专注于评估机器理解中使用常识知识的能力。该数据集包含约2,100个叙述文本和约14,000个关于这些文本的问题，特别关注日常活动的故事，如去电影院或园艺，并要求使用常识知识，特别是脚本知识来回答问题。数据集通过众包收集，经过手动验证和过滤，确保高质量。MCScript不仅为自然语言理解社区提供了挑战性的测试案例，还作为2018年SemEval的一个共享任务的基础，旨在评估和推动机器理解技术的发展。

MCScript is a large-scale dataset developed by Saarland University in Germany, focusing on evaluating machines' ability to utilize common-sense knowledge for machine comprehension. The dataset contains approximately 2,100 narrative texts and around 14,000 questions related to these texts, with a particular emphasis on stories of daily activities such as going to the cinema or gardening, and requires the use of common-sense knowledge, especially script knowledge, to answer the questions. It was collected via crowdsourcing, and then manually validated and filtered to ensure high data quality. Beyond offering challenging test cases for the natural language understanding community, MCScript also served as the foundation for a shared task at SemEval 2018, which aims to evaluate and promote the advancement of machine comprehension technologies.