zhibei1204/DiagramQG

DiagramQG是一个专注于科学图表问题生成的教育数据集，包含19,475个独特问题、8,372个图表以及44,472个（目标与概念文本约束、图表、问题）的组合。数据集覆盖了4个学科（自然科学、地球科学、应用科学和社会科学）、15门课程和169个概念。数据收集过程分为四个阶段：初始数据收集、组织、标注和质量保证。数据集的结构按学科、课程和概念进行层次化组织。

The DiagramQG dataset is a comprehensive educational dataset focused on generating concept-focused questions from scientific diagrams. It contains 19,475 unique questions, 8,372 diagrams, and 44,472 combinations of (target & concept text constraint, diagram, question). The dataset covers four main subject areas: Natural Science, Earth Science, Applied Science, and Social Science, organized hierarchically into subjects, courses, and concepts. The data collection process includes initial data gathering, classification, annotation, and quality assurance. Unique challenges of the dataset include domain-specific knowledge requirements, long-tail distribution, and high information density.