ProPara

该数据集是由众包方式收集的描述流程的段落，用于测试检索类比算法。每个提示由1至6名工作者完成，训练集包含390个段落。数据规模约为390个段落，大约76,000个可能的段落对。该任务旨在寻找描述流程的段落之间的类比关系。

This dataset comprises workflow-descriptive paragraphs collected through crowdsourcing for evaluating analogy retrieval algorithms. Each annotation prompt is completed by 1 to 6 crowd workers. The training set includes 390 paragraphs, yielding approximately 76,000 potential paragraph pairs. The core task of this work is to identify analogical connections between these workflow-descriptive paragraphs.