FUNSD
收藏资源简介:
FUNSD是由瑞士联邦理工学院信号处理实验室5创建的数据集,专注于噪声扫描文档中的表单理解。该数据集包含199个真实、全标注的扫描表单,这些表单在外观上具有广泛的变化,适用于文本检测、光学字符识别、空间布局分析和实体标注/链接等任务。数据集的创建过程采用自底向上的方法进行标注,确保了数据集在文档理解任务中的多样性和实用性。FUNSD数据集的应用领域主要集中在自动化信息提取和结构化,旨在解决从扫描文档中提取和理解信息的问题。
FUNSD is a dataset developed by Signal Processing Laboratory 5 of ETH Zurich (Swiss Federal Institute of Technology Zurich), focusing on form understanding in noisy scanned documents. The dataset comprises 199 real, fully annotated scanned forms with wide visual variability, which is applicable to tasks including text detection, optical character recognition (OCR), spatial layout analysis, and entity annotation/linking. A bottom-up annotation approach was adopted during the dataset's creation, ensuring its diversity and practicality for document understanding tasks. The primary application areas of the FUNSD dataset center on automated information extraction and structuring, aiming to address the challenges of extracting and understanding information from scanned documents.

- FUNSD数据集首次发表,由Gomez-Adorno等人提出,旨在为表单理解任务提供一个标准化的数据集。
- FUNSD数据集在多个自然语言处理和计算机视觉会议上被广泛引用,成为表单理解领域的重要基准。
- FUNSD数据集的应用扩展到多模态学习领域,研究人员开始探索如何结合文本、图像和其他模态信息来提升表单理解的效果。
- FUNSD数据集的改进版本发布,增加了更多的表单样本和多样化的数据类型,进一步推动了表单理解技术的发展。
- 1FUNSD: A Dataset for Form Understanding in Noisy Scanned DocumentsUniversity of Antwerp · 2019年
- 2LayoutLM: Pre-training of Text and Layout for Document Image UnderstandingMicrosoft Research Asia · 2019年
- 3DocFormer: End-to-End Transformer for Document UnderstandingIndian Institute of Technology Madras · 2021年
- 4BERT for Form Understanding: A Case Study on FUNSD DatasetUniversity of California, Irvine · 2020年
- 5FormNet: Structural Encoding beyond Sequential Modeling in Form Document Information ExtractionUniversity of California, Irvine · 2020年



