https://doi.org/10.1140/epjds/s13688-026-00618-5
Research
Combining large language and deep learning models for media text classification to perform supply chain due diligence risk assessments
1
Reliable, Secure and Connected Systems Unit, Luxembourg Institute of Science of Technology, 5 Av. des Hauts-Fourneaux, 4362, Esch-sur-Alzette, Luxembourg
2
AI Research & Development, CURE SA, 3 Schaffmill, 6778, Gréiwemaacher, Luxembourg
3
scrioo SARL, 2 Kofferschmattsgaessel, 6778, Gréiwemaacher, Luxembourg
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
b
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
31
March
2025
Accepted:
5
January
2026
Published online:
16
January
2026
Abstract
New legislation requires companies to regularly perform environmental, social, and governance (ESG) risk assessments for their suppliers. This is typically done using public risk indicators for countries and industries in which the suppliers operate. However, the approach often does not represent the actual risk stemming from a particular supplier accurately. Moreover, risk indicators are usually only updated annually and do not reflect current developments. Therefore, big text data collected from media monitoring on the suppliers can augment these risk indicators to provide more accurate and timely risk assessments. Although, using media texts for this purpose is challenging. It usually has a low signal-to-noise ratio and requires reliable complex text interpretation to be usable. Against this background, we propose a hybrid approach of large language models and deep learning classifiers to process media texts for automated ESG risk assessment. A set of 17 classifiers is investigated and tested against the direct usage of large language models for classification. The performance is evaluated in a Monte Carlo experiments on four distinct datasets.
Key words: Big data / Generative AI / Hybrid machine learning / Process automation / Risk management
Handling Editor: Francesco Gullo
Djamel Khadraoui, Stéphane Cortina, Cathrine Aeckerle-Willems and Felix Viroli contributed equally to this work.
© The Author(s) 2026
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
