REGLAT at MAHED Shared Task: A Hybrid Ensemble-Based System for Arabic Hate Speech Detection

Résumé

Hope and hate speech detection in natural language processing addresses the challenge of identifying social media content within the fastpaced environment of online platforms.Hopeful speech that promotes supportive and inclusive language plays a crucial role in counteracting online toxicity, whereas hate speech poses threats and challenges to society.This paper focuses on text-based Arabic hate and hope speech detection, demonstrating the system submitted by the REGLAT team to the MAHED shared task held in conjunction with ArabicNLP 2025.The proposed system employs an ensemble-based model that combines a TF-IDF + Logistic Regression classifier with a fine-tuned AraBERTv2 model as baselines.A majority voting approach is then applied to aggregate the predictions.The proposed model reported an F1 score of 0.58.These promising results are notable given the simplicity of the system's architecture, and they highlight the potential of our approach for improving the performance of this task.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0