Request Correction

Viva_Palestine at StanceNakba Shared Task: Actor and Topic-Aware Stance Detection in Public Discourse

Recent research has increasingly focused on user-generated content to clarify opinions expressed in social media discourse. The Actor and Topic-Aware Stance Detection in Public Discourse challenge encourages research on stance detection in polarized social media discourse on the Palestinian–Israeli conflict. The challenge comprises two subtasks: one for actor-level alignments and the other for cross-topic generalization patterns. The StanceNakba2026 task includes two subtasks: (A) Actor-Level Stance Detection in English and (B) Cross-Topic Stance Detection in Arabic. Our team participated in both subtasks with the name "Viva_Palestine". In Subtask A, the proposed method is based on the Bert-Base-Uncased model and achieved a Macro F1-score of 0.9190, placing 6th out of 13 teams. In Subtask B, the proposed method is based on the MARBERT model and achieved a Macro F1-score of 0.8724 (the top rank in the leaderboard), placing first out of 10 teams. These results show that the proposed modelling method performs well for both entity-specific stance alignment and strong cross-topic generalization.