Ryoo72/DocStruct4M_ip

这是一个包含文本和图像路径的数据集，用于视觉句子融合任务。它由训练集和验证集组成，文本数据包含索引、文本内容和类型，以及角色信息。图像不是直接嵌入的，而是通过路径引用，需要从原始数据集下载和解压图像文件来获取。

This is a dataset containing text and image paths for visual sentence fusion tasks. It consists of a training set and a validation set, with text data including index, text content, type, and role information. Images are not embedded directly but are referenced by paths, requiring downloading and extracting image files from the original dataset.