An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect
Researchers have developed a novel hybrid framework for detecting rumors in the Algerian dialect of Arabic, a challenging task due to informal language and limited resources. The framework combines transformer embeddings with classical classifiers, achieving an F1-score of 0.84. Domain-specific pre-training proved more critical than model size, with models trained on social media data outperforming those trained on formal Arabic corpora. AI
IMPACT This research offers a potential solution for rumor detection in low-resource dialects, which could be adapted for other languages facing similar challenges.