TextOCR

该数据集是一个为文本识别和编辑任务设计的真实世界数据集，它是用于指令调整训练的一部分。具体来说，该数据集与其他数据集结合，共同构成了约130万个示例的指令调整训练集，旨在提升文本渲染和编辑任务的性能。

This is a real-world dataset designed for text recognition and editing tasks, and it serves as a component of instruction tuning training. Specifically, when combined with other datasets, it jointly forms an instruction tuning training set containing approximately 1.3 million examples, with the goal of enhancing the performance of text rendering and editing tasks.