A Scoping Review of Arabic Natural Language Processing for Mental Health

Area of Science:

• Computational linguistics
• Mental health informatics
• Arabic natural language processing

Background:

• Mental health disorders are a significant global health issue.
• Natural Language Processing (NLP) offers tools for analyzing text data to identify mental health challenges.
• Arabic NLP in mental health research requires systematic review.

Purpose of the Study:

• To identify Arabic NLP techniques used in mental health research.
• To determine the mental health conditions addressed by these techniques.
• To evaluate the effectiveness of NLP in detecting and predicting mental health conditions in Arabic text.

Main Methods:

• Scoping review conducted following the PRISMA-ScR framework.
• Systematic literature search across PubMed, ScienceDirect, IEEE Xplore, and Google Scholar.
• Inclusion of studies on Arabic NLP for mental health in peer-reviewed publications.

Main Results:

• Various NLP techniques were identified, including Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Recurrent Neural Network (RNN), and transformer models (AraBERT, MARBERT).
• Predominant focus on detecting depression and suicidality from Arabic social media data.
• Transformer-based models (AraBERT, MARBERT) achieved superior performance, with accuracies up to 99.3% and 98.3% respectively.

Conclusions:

• NLP, especially advanced transformer models, holds significant potential for addressing mental health issues in Arabic-speaking populations.
• Transformer models demonstrate superior accuracy and insight compared to traditional ML and RNNs.
• Continued research is vital to advance the field and validate findings.