Sakinah-AI: Optimized CAMeL-BERT for Arabic Mental Health Question Classification

This repository contains the official fine-tuned model Sakinah-AI-CAMEL-BERT-Optimized, our submission to the MentalQA 2025 Shared Task (Track 1).

By: Fatimah Emad Elden & Mumina Abukar

Cairo University & The University of South Wales

---

Model Description

This model is a fine-tuned version of CAMeL-Lab/bert-base-arabic-camelbert-mix-sentiment for multi-label classification of Arabic questions related to mental health. It was trained on the AraHealthQA dataset.

Our approach involved a comprehensive hyperparameter search using the Optuna framework to find the optimal configuration. To address the inherent class imbalance in the dataset, the model was trained using a custom Focal Loss function. This strategy proved highly effective, making this model our best-performing fine-tuned system, achieving a Weighted F1-score of 0.597 on the official blind test set.

The model predicts one or more of the following labels for a given question:

• A: Diagnosis (Interpreting symptoms)
• B: Treatment (Seeking therapies or medications)
• C: Anatomy and Physiology (Basic medical knowledge)
• D: Epidemiology (Course, prognosis, causes of diseases)
• E: Healthy Lifestyle (Diet, exercise, mood control)
• F: Provider Choices (Recommendations for doctors)
• Z: Other (Does not fit other categories)

---

🚀 How to Use

You can use this model directly with the transformers library pipeline for text-classification.

from transformers import pipeline

# Load the classification pipeline
classifier = pipeline(
    "text-classification",
    model="FatimahEmadEldin/Sakinah-AI-CAMEL-BERT-Optimized",
    return_all_scores=True # Set to True for multi-label output
)

# Example question in Arabic
question = "ما هي أعراض الاكتئاب وكيف يمكن علاجه؟"
# (Translation: "What are the symptoms of depression and how can it be treated?")

results = classifier(question)

# --- Post-processing to get final labels ---
# The optimal threshold found during tuning was ~0.25
threshold = 0.246
predicted_labels = [item['label'] for item in results[0] if item['score'] > threshold]

print(f"Question: {question}")
# Expected output for this example would likely include 'Diagnosis' and 'Treatment'
print(f"Predicted Labels: {predicted_labels}")
# [['A', 'B']]

⚙️ Training Procedure

This model was fine-tuned using a rigorous hyperparameter optimization process.

Hyperparameters

The best hyperparameters found by Optuna and used for this model are:

Frameworks

• PyTorch
• Hugging Face Transformers
• Optuna

📊 Evaluation Results

The model was evaluated on the blind test set provided by the MentalQA organizers.

Final Test Set Scores

Per-Label Performance (Test Set)

              precision    recall  f1-score   support
           A       0.61      0.96      0.75        84
           B       0.56      0.98      0.72        85
           C       0.14      0.40      0.21        10
           D       0.25      0.91      0.39        34
           E       0.34      0.53      0.41        38
           F       0.07      0.33      0.11         6
           Z       0.00      0.00      0.00         3

   micro avg       0.42      0.85      0.57       260
   macro avg       0.28      0.59      0.37       260
weighted avg       0.47      0.85      0.60       260
 samples avg       0.44      0.88      0.56       260

📜 Citation

If you use our work, please cite our paper:

@inproceedings{elden2025sakinahai,
    title={{Sakinah-AI at MentalQA: A Comparative Study of Few-Shot, Optimized, and Ensemble Methods for Arabic Mental Health Question Classification}},
    author={Elden, Fatimah Emad and Abukar, Mumina},
    year={2025},
    booktitle={Proceedings of the MentalQA 2025 Shared Task},
    eprint={25XX.XXXXX},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}