ChartBank
收藏资源简介:
ChartBank是由香港科技大学(广州)和香港科技大学团队创建的大规模数据集,旨在通过整合视觉和文本信息来提升图表理解任务的性能。该数据集包含约750万条样本,涵盖了多种图表相关任务,如OCR感知数据提示、视觉提示和图表到表格的指令跟随任务。数据集的创建过程包括从图表中提取文本信息,并将其与视觉元素对齐,以生成有效的联合表示。ChartBank的应用领域主要集中在图表问答、图表到文本转换等任务,旨在解决现有模型在处理图表时未能充分利用文本信息的问题。
ChartBank is a large-scale dataset developed by teams from The Hong Kong University of Science and Technology (Guangzhou) and The Hong Kong University of Science and Technology. It aims to improve the performance of chart understanding tasks by integrating visual and textual information. The dataset contains approximately 7.5 million samples, covering a wide range of chart-related tasks including OCR-aware data prompting, visual prompting, and chart-to-table instruction following tasks. The dataset construction process involves extracting textual information from charts and aligning it with visual elements to generate effective joint representations. The primary application areas of ChartBank focus on tasks such as chart question answering and chart-to-text conversion, aiming to address the issue that existing models fail to fully leverage textual information when processing charts.

- 1AskChart: Universal Chart Understanding through Textual Enhancement香港科技大学(广州)和香港科技大学 · 2024年



