CS316 · Mental Health AI Project

Embedding-Based Longitudinal Monitoring for Depression and Anxiety Severity (Arabic NLP)

This project applies Natural Language Processing and Machine Learning to analyze Arabic textual narratives related to mental health.

The system evaluates depression and anxiety severity using embedding-based representations and classical classifiers. It extends beyond static prediction by integrating a trajectory-based alert framework for early risk detection.

Developed for CS316, Artificial Intelligence and Data Science (Semester 252) at Prince Sultan University. The project follows a research-oriented and responsible AI methodology.

View Project Overview Read IEEE Paper Draft

---

What This Project Does

• Severity Classification

  ---

  Dual-model architecture predicts depression and anxiety severity independently on a 0 to 3 scale using embedding-based feature representations and SVM classifiers.

• Longitudinal Monitoring

  ---

  Instead of one-time screening, the system tracks mental health trajectories across multiple time points to detect meaningful changes.

• Predictive Alert Mechanism

  ---

  Three alert categories identify:

  • Sustained high severity
  • Gradual worsening trends
  • Sudden severity spikes

  This supports early awareness and intervention research.

• Responsible AI Framework

  ---

  Designed as a decision-support research tool, not a diagnostic system. Emphasizes privacy, synthetic data usage, transparency, and fairness.

---

Technical Foundation

• Arabic NLP

  ---

  Uses multilingual embedding models to represent Arabic text in high-dimensional semantic space.

  Methodology

• Machine Learning

  ---

  Support Vector Machines with RBF kernel trained using stratified sampling for robust multi-class classification.

  Model Details

• Experimental Results

  ---

  High classification accuracy with strong macro-averaged performance and high AUC scores.

  Results and Analysis

• Deployment

  ---

  Demonstrates how the system can be used for research experiments and controlled evaluation.

  Deployment Guide

---

Project Summary

This project investigates how embedding-based NLP models combined with classical machine learning can support structured and longitudinal assessment of mental health indicators in Arabic text.

Rather than replacing clinical evaluation, the system functions as a research-driven early warning monitoring tool. It integrates semantic embeddings, multi-class severity prediction, and temporal alert logic within a responsible AI framework.

The emphasis is on:

• Technical rigor
• Ethical AI design
• Reproducibility
• Transparent evaluation

---