Multilingual Propaganda Detection: Exploring Transformer-Based Models mBERT, XLM-RoBERTa, and mT5

Mohamed Ibrahim Ragab, Ensaf Hussein Mohamed, Walaa Medhat


Abstract
This research investigates multilingual propaganda detection by employing transformer-based models, specifically mBERT, XLM-RoBERTa, and mT5. The study utilizes a balanced dataset from the BiasFigNews corpus, annotated for propaganda and bias across five languages. The models were finely tuned to generate embeddings for classification tasks. The evaluation revealed mT5 as the most effective model, achieving an accuracy of 99.61% and an F1-score of 0.9961, followed by mBERT and XLM-RoBERTa with accuracies of 92% and 91.41%, respectively. The findings demonstrate the efficacy of transformer-based embeddings in detecting propaganda while also highlighting challenges in subtle class distinctions. Future work aims to enhance cross-lingual adaptability and explore lightweight models for resource-constrained settings.
Anthology ID:
2025.nakbanlp-1.9
Volume:
Proceedings of the first International Workshop on Nakba Narratives as Language Resources
Month:
January
Year:
2025
Address:
Abu Dhabi
Editors:
Mustafa Jarrar, Habash Habash, Mo El-Haj
Venues:
NakbaNLP | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
75–82
Language:
URL:
https://aclanthology.org/2025.nakbanlp-1.9/
DOI:
Bibkey:
Cite (ACL):
Mohamed Ibrahim Ragab, Ensaf Hussein Mohamed, and Walaa Medhat. 2025. Multilingual Propaganda Detection: Exploring Transformer-Based Models mBERT, XLM-RoBERTa, and mT5. In Proceedings of the first International Workshop on Nakba Narratives as Language Resources, pages 75–82, Abu Dhabi. Association for Computational Linguistics.
Cite (Informal):
Multilingual Propaganda Detection: Exploring Transformer-Based Models mBERT, XLM-RoBERTa, and mT5 (Ragab et al., NakbaNLP 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.nakbanlp-1.9.pdf