Leveraging Multilingual NLP Models for Mental Health Detection in Arabic Texts: A Critical Overview

Background : : Mental health issues such as depression, anxiety, and suicidal ideation are increasingly visible in digital communications, particularly within Arabic-speaking populations. Natural language processing (NLP) has shown promise in enabling early detection and automated analysis of such concerns, but Arabic NLP remains technically and linguistically underdeveloped.Objective: This review examines recent advances in Arabic mental health NLP, with a focus on multilingual and transformer-based models. It aims to synthesize findings from studies using Arabictet detect or classify mental health conditions, while evaluating the tools, datasets, and ethical considerations involved.Methods: A scoping literature review was conducted across major scientific databases, including PubMed, Scopus, and Google Scholar, covering studies from 2016 to 2024. Only peer-reviewed works using machine learning or deep learning techniques on Arabic data related to mental health were included.Findings: Twenty-two studies were selected, primarily employing models such as AraBERT, MARBERT, BiLSTM, and SVM on Arabic tweets, clinical notes, and forum posts. Depression was the most frequently studied condition. While accuracy rates exceeded 80% in many models, limitations included data scarcity, dialectal variability, and inconsistent annotation standards.Conclusion: Arabic mental health NLP is a growing but fragmented field. Future work should focus on corpus standardization, inclusion of dialects, and ethical frameworks to ensure culturally sensitive and effective deployment of AI-driven tools for psychological support.