Prix Nakano "Best Paper Award" à DAS 2022 pour Thibault Douzon

Le 25 mai 2022, Thibault Douzon, doctorant de l’équipe Imagine en partenariat avec l'entreprise Esker, a reçu le prix Nakano qui récompense le « Best Paper » pour son article intitulé « Improving Information Extraction in Business Documents with Specific Pre-Training ». Son travail réalisé sous la direction de Stefan Duffner, Christophe Garcia et Jérémy Espinas a été présenté à la 15e édition du workshop international Document Analysis Systems (DAS) à La Rochelle.

Résumé (en anglais) :

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.

Lien de la conférence :