Learning to Extract Semantic Structure from Documents Using Multimodal Fully Convolutional Neural Network

TL;DR

This paper introduces a multimodal fully convolutional neural network that combines visual and textual information to accurately extract semantic structures from document images, utilizing synthetic pretraining and semi-supervised fine-tuning.

Contribution

It presents a novel end-to-end multimodal network architecture and an efficient synthetic data generation method for improved document semantic structure extraction.

Findings

Multimodal approach significantly outperforms visual-only methods.

Synthetic pretraining enhances model performance.

Semi-supervised fine-tuning on real documents improves accuracy.

Abstract

We present an end-to-end, multimodal, fully convolutional network for extracting semantic structures from document images. We consider document semantic structure extraction as a pixel-wise segmentation task, and propose a unified model that classifies pixels based not only on their visual appearance, as in the traditional page segmentation task, but also on the content of underlying text. Moreover, we propose an efficient synthetic document generation process that we use to generate pretraining data for our network. Once the network is trained on a large set of synthetic documents, we fine-tune the network on unlabeled real documents using a semi-supervised approach. We systematically study the optimum network architecture and show that both our multimodal approach and the synthetic data pretraining significantly boost the performance.