letxbe/BoundingDocs
收藏官方服务:
资源简介:
BoundingDocs是一个用于文档问答的大型空间注释数据集,包含48,151个文档,跨越多种文档类型,支持8种语言。它将信息提取任务转化为问答任务,适用于训练和评估大型语言模型,并提供精确的边界框注释以提高文档理解。
BoundingDocs is a large-scale spatially-annotated dataset for Document Question Answering (QA), encompassing 48,151 documents across various types and supporting 8 languages. It transforms Information Extraction tasks into QA tasks, suitable for training and evaluating Large Language Models, with precise bounding box annotations to enhance document comprehension.
提供机构:
letxbe


