Improving Information Extraction on Business Documents with Specific Pre-Training Tasks
This work addresses the challenge of extracting complex information from business documents like receipts and invoices, representing an incremental improvement over existing methods.
The paper tackled the problem of insufficient pre-training tasks for extracting information from business documents by introducing two new pre-training tasks and a post-processing algorithm, resulting in improved F1 scores from 93.88 to 95.50 on public datasets and from 84.35 to 84.84 on private datasets.
Transformer-based Language Models are widely used in Natural Language Processing related tasks. Thanks to their pre-training, they have been successfully adapted to Information Extraction in business documents. However, most pre-training tasks proposed in the literature for business documents are too generic and not sufficient to learn more complex structures. In this paper, we use LayoutLM, a language model pre-trained on a collection of business documents, and introduce two new pre-training tasks that further improve its capacity to extract relevant information. The first is aimed at better understanding the complex layout of documents, and the second focuses on numeric values and their order of magnitude. These tasks force the model to learn better-contextualized representations of the scanned documents. We further introduce a new post-processing algorithm to decode BIESO tags in Information Extraction that performs better with complex entities. Our method significantly improves extraction performance on both public (from 93.88 to 95.50 F1 score) and private (from 84.35 to 84.84 F1 score) datasets composed of expense receipts, invoices, and purchase orders.