reading order

AntTech

Nov 15, 2023 · Artificial Intelligence

Reading Order Matters: Information Extraction from Visually‑rich Documents by Token Path Prediction

The paper identifies reading‑order disorder as a critical obstacle in visually‑rich document information extraction, proposes a Token Path Prediction model with grid‑label formulation, introduces re‑annotated FUNSD‑r and CORD‑r datasets, and demonstrates SOTA performance on NER, entity linking, and reading‑order prediction tasks.