Jointly Learning Span Extraction and Sequence Labeling for Information Extraction from Business Documents

TL;DR

This paper presents a novel end-to-end model that jointly learns span extraction and sequence labeling to improve information extraction from business documents, especially long and sparse ones.

Contribution

It introduces a unified model combining span extraction and sequence labeling, enhancing extraction accuracy and speed over traditional methods.

Findings

Achieves promising results on four business datasets in English and Japanese.

Significantly faster than traditional span-based extraction methods.

Demonstrates effectiveness in handling long, sparse documents.

Abstract

This paper introduces a new information extraction model for business documents. Different from prior studies which only base on span extraction or sequence labeling, the model takes into account advantage of both span extraction and sequence labeling. The combination allows the model to deal with long documents with sparse information (the small amount of extracted information). The model is trained end-to-end to jointly optimize the two tasks in a unified manner. Experimental results on four business datasets in English and Japanese show that the model achieves promising results and is significantly faster than the normal span-based extraction method. The code is also available.