MentalMARBERT: Domain-Adaptive Pre-training and Two-Stage Fine-Tuning for Arabic Mental Health Disorders Detection
Researchers have developed a novel framework, MentalMARBERT, to improve the detection of mental health disorders from Arabic social media text. The approach involves domain-adaptive pre-training of existing Arabic language models like MARBERT on a large corpus of unlabeled mental health tweets. This adapted model was then fine-tuned using a two-stage classification architecture, achieving a macro-F1 score of 0.861 and an accuracy of 0.877 on a newly constructed dataset of over 50,000 tweets. The study highlights the effectiveness of specialized pre-training and hierarchical classification for this challenging NLP task. AI
IMPACT This research advances Arabic NLP capabilities, potentially improving mental health support accessibility in Arabic-speaking communities.