FormNet: Structural Encoding beyond Sequential Modeling in Form Document Information Extraction

TL;DR

FormNet introduces a structure-aware sequence model that leverages spatial relationships and super-tokens to improve form document information extraction, outperforming existing methods with less data.

Contribution

It proposes Rich Attention and Super-Tokens to better encode spatial and local syntactic information in form documents, addressing serialization challenges.

Findings

Outperforms existing methods on CORD, FUNSD, and Payment benchmarks.

Achieves state-of-the-art performance with a more compact model and less pre-training data.

Effectively captures spatial relationships and local syntax in form documents.

Abstract

Sequence modeling has demonstrated state-of-the-art performance on natural language and document understanding tasks. However, it is challenging to correctly serialize tokens in form-like documents in practice due to their variety of layout patterns. We propose FormNet, a structure-aware sequence model to mitigate the suboptimal serialization of forms. First, we design Rich Attention that leverages the spatial relationship between tokens in a form for more precise attention score calculation. Second, we construct Super-Tokens for each word by embedding representations from their neighboring tokens through graph convolutions. FormNet therefore explicitly recovers local syntactic information that may have been lost during serialization. In experiments, FormNet outperforms existing methods with a more compact model size and less pre-training data, establishing new state-of-the-art performance on CORD, FUNSD and Payment benchmarks.