UniBO at CheckThat! 2024 Multi-lingual and Multi-label Persuasion Technique Detection in News with Data Augmentation and Sequence-Token Classifiers

We explored persuasion technique detection in multilingual news as part of the CheckThat! Lab Task 3. Our pipeline comprises two parts. The first part is a data augmentation module, which uses a BERT-based model fine-tuned for word-alignment to project labels from source texts to machine-translated target texts. The second one is a persuasion technique classification module, leveraging two fine-tuned BERT-based models: a sequence classifier for detecting sentences containing persuasion techniques and a set of 23 token-level classifiers for specific techniques. Our approach sis competitive in all language settings, showing hints of crosslingual transfer. Despite the research efforts in this direction, detecting persuasion techniques, especially across languages, remains challenging due to their implicit and subtle nature.

Sep 2, 2024 4:40 PM — Sep 12, 2024 4:30 PM