A Dataset of Alt Texts from HCI Publications

本数据集由华盛顿大学和艾伦人工智能研究所创建，专注于从人机交互（HCI）出版物中提取的替代文本（alt text），旨在帮助视障读者理解科学论文中的数据可视化内容。数据集包含3386条作者编写的替代文本，主要来源于HCI和可访问性出版物，重点关注图表、图和绘图的替代文本。创建过程中，研究者从超过25,000篇出版物中提取了近3,400条有效替代文本。该数据集的应用领域包括开发工具和模型，以辅助作者编写更详细的替代文本，并研究图像理解模型和众包技术如何更有效地生成科学领域的图表替代文本。

This dataset was created by the University of Washington and the Allen Institute for Artificial Intelligence, focusing on alt text extracted from Human-Computer Interaction (HCI) publications with the goal of assisting visually impaired readers in understanding data visualizations in scientific papers. The dataset contains 3,386 author-written alt texts, primarily sourced from HCI and accessibility publications, with a focus on alt texts for charts, figures, and plots. During its creation, researchers extracted nearly 3,400 valid alt texts from over 25,000 publications. Potential applications of this dataset include developing tools and models to help authors compose more detailed alt texts, as well as investigating how image understanding models and crowdsourcing technologies can more effectively generate alt texts for scientific charts.