LayoutMask: Enhance Text-Layout Interaction in Multi-modal Pre-training for Document Understanding

TL;DR

LayoutMask is a novel multi-modal pre-training model that enhances text-layout interaction in visually-rich document understanding, achieving state-of-the-art results across various document analysis tasks.

Contribution

It introduces a new layout input method and two pre-training objectives to improve text-layout fusion in document understanding models.

Findings

Achieves state-of-the-art performance on form understanding.

Improves robustness in receipt understanding.

Enhances document image classification accuracy.

Abstract

Visually-rich Document Understanding (VrDU) has attracted much research attention over the past years. Pre-trained models on a large number of document images with transformer-based backbones have led to significant performance gains in this field. The major challenge is how to fusion the different modalities (text, layout, and image) of the documents in a unified model with different pre-training tasks. This paper focuses on improving text-layout interactions and proposes a novel multi-modal pre-training model, LayoutMask. LayoutMask uses local 1D position, instead of global 1D position, as layout input and has two pre-training objectives: (1) Masked Language Modeling: predicting masked tokens with two novel masking strategies; (2) Masked Position Modeling: predicting masked 2D positions to improve layout representation learning. LayoutMask can enhance the interactions between text and layout modalities in a unified model and produce adaptive and robust multi-modal representations for downstream tasks. Experimental results show that our proposed method can achieve state-of-the-art results on a wide variety of VrDU problems, including form understanding, receipt understanding, and document image classification.