ROPE: Reading Order Equivariant Positional Encoding for Graph-based Document Information Extraction

TL;DR

This paper introduces ROPE, a novel positional encoding method that enhances graph neural networks' ability to understand reading order in document images, significantly improving entity extraction accuracy.

Contribution

ROPE provides a new reading order-aware positional encoding for GCNs, improving document information extraction by capturing sequential word presentation.

Findings

ROPE improves GCN performance by up to 8.4% F1-score.

ROPE effectively captures reading order in document graphs.

Enhanced entity extraction on FUNSD and payment datasets.

Abstract

Natural reading orders of words are crucial for information extraction from form-like documents. Despite recent advances in Graph Convolutional Networks (GCNs) on modeling spatial layout patterns of documents, they have limited ability to capture reading orders of given word-level node representations in a graph. We propose Reading Order Equivariant Positional Encoding (ROPE), a new positional encoding technique designed to apprehend the sequential presentation of words in documents. ROPE generates unique reading order codes for neighboring words relative to the target word given a word-level graph connectivity. We study two fundamental document entity extraction tasks including word labeling and word grouping on the public FUNSD dataset and a large-scale payment dataset. We show that ROPE consistently improves existing GCNs with a margin up to 8.4% F1-score.