AraNLP at MAHED 2025 Shared Task: Using AraBERT for Text-based Hate and Hope Speech Classification
Year: 2025
Type: conference-paper
Abstract: The AraNLP system is designed for the MAHED Shared Task 2025 on text-based hate and hope classification in Arabic.The challenge was divided into three subtasks:(1) Text-based Hate and Hope Speech Classification, (2) Emotion, Offensive, and Directed Hate Detection (Multitask), and (3) Multimodal Hateful Meme Detection.The AraNLP system based on the AraBERT model achieved Macro F1-score 65% for Sub-task 1 and the results are published in the leaderboard, with rank 20.After submitting the results, the proposed model was updated to improve its performance and achieved Macro F1-score 70%, this makes the AraNLP system nearly equivalent to rank 4 in the leaderboard.
Authors: Wafaa S. El-Kassas, Enas A. Hakim Khalil
Institutions: Egyptian e-Learning University, National Research Centre
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4416037803