OPENPI

OPENPI是由艾伦人工智能研究所创建的一个大规模高质量数据集，包含29,928个状态变化，覆盖了4,050个句子，来源于810个WikiHow.com上的实际操作段落。该数据集通过众包方式创建，旨在捕捉开放领域程序文本中的实体状态变化。数据集内容涵盖多种实体和属性，如位置、温度等，创建过程中使用了预定义的51个WordNet属性，并鼓励众包工作者自由描述状态变化。OPENPI的应用领域包括自动执行生物实验、烹饪食谱和日常活动等，旨在解决机器理解文本中隐含动作效果的问题。

OPENPI is a large-scale high-quality dataset created by the Allen Institute for Artificial Intelligence. It comprises 29,928 state changes spanning 4,050 sentences, which are sourced from 810 practical instructional passages on WikiHow.com. Built via a crowdsourcing approach, this dataset aims to capture entity state changes in open-domain procedural texts. The dataset covers a diverse range of entities and attributes such as location, temperature, and more. During its development, 51 pre-defined WordNet attributes were utilized, and crowdsourcing workers were encouraged to freely describe state changes. The application scenarios of OPENPI include automated biological experiments, cooking recipes, daily activities and more, with the goal of solving the problem of machines comprehending the implicit action effects conveyed in texts.