QuranicJar 🕌✨

QuranicJar is a SwiftUI-based iOS application that predicts a user's emotion based on their input and provides a relevant Quranic verse for guidance and comfort. The app uses a Flask-based backend API powered by fine-tuned Natural Language Processing (NLP) BERT and RoBERTa models for emotion classification.

Table of Contents

• Features
• File Description
• How it Works
• How to Run the Project
• Future Enhancements
• Contributions
• Acknowledgments
• License

Features

• Emotion Prediction: Users can input their feelings, and the app predicts their emotion using a backend API.
• Supported Emotions: As for now, there are 4 supported emotions which are anger, fear, joy and sadness.
• Quranic Verse Recommendation: Based on the predicted emotion, the app provides a relevant Quranic verse for guidance.
• Bookmark Verse: Bookmark and filter your saved verses based on emotions.
• Search Verse: Search verses according to emotion types, Surah's name or translation keywords.
• Ensemble Model: Combines predictions from BERT and RoBERTa (NLP) models for improved accuracy.
• Loading Indicator: Displays a progress view while the emotion prediction is being processed.

File Descriptions

1. QuranJarApp.swift:

   • The main entry point of the iOS app.

2. ContentView.swift:

   • The main screen where users can input their feelings, predict emotions, and view Quranic verses.
   • Sends API requests to the Flask backend for emotion prediction.

3. WelcomeView.swift:

   • The initial screen where users are greeted and asked to enter their name.

4. app.py:

   • Flask-based backend API for emotion prediction.
   • Uses fine-tuned BERT and RoBERTa models to classify emotions.
   • Maps predicted emotions to Quranic verses using a dataset.

5. train_bert.py,train_roberta.py,train_distilbert.py

   • Fine tune the pretrain models to the specific dataset

6. test_singlemodel.py,test_ensemblemodel.py

   • Testing the fine-tuned model

7. quran_emotions.csv:

   • A dataset containing Quranic verses mapped to specific emotions.

8. model/:

   • Contains the fine-tuned BERT and RoBERTa models and their tokenizers.

Backend API

The backend API is built using Flask and provides an endpoint for emotion prediction.

API Endpoint

POST /predict

Request Format

The API expects a JSON payload with the following structure:

{
  "text": "User's input text"
}

Response Format

The API returns a JSON response with the following structure:

{
  "predicted_emotion": "Emotion name",
  "probabilities": [0.1, 0.2, 0.5, 0.2],
  "quranic_verse": "Relevant Quranic verse (Surah X, Verse Y)"
}

How It Works

1. The input text is processed by both the fine-tuned BERT and RoBERTa models.
2. The predictions from both models are combined using an ensemble method.
3. The predicted emotion is mapped to a relevant Quranic verse using the quran_emotions.csv dataset.

---

How to Run the Project

Prerequisites

• Python 3.8 or later
• Flask
• PyTorch
• Transformers library
• pandas

Backend Setup

1. Clone the repository:

   git clone https://github.com/your-username/QuranicJar.git
   cd QuranicJar

2. Install the required Python packages:

   pip install -r requirements.txt

3. Ensure the quran_emotions.csv dataset is in the dataset/ directory.

4. Run the Flask app:

   python app.py

   The backend will be available at http://0.0.0.0:3000.

iOS App Setup

1. Open the swift directory in Xcode.
2. Run the app on a simulator or a physical device.
3. Ensure the backend API is running and accessible from the device.

Requirements

Backend

• Python 3.8 or later
• Flask
• PyTorch
• Transformers
• pandas

iOS App

• iOS 14.0 or later
• Xcode 12.0 or later
• Swift 5.0 or later

Future Enhancements

• Add support for more emotions and Quranic verses.
• Improve the emotion prediction model with additional training data.
• Enhance the UI/UX of the iOS app.

Contributions

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgments

• Quran Dataset
• Emotionally-Labeled Quran Verses (ELQV)
• BERT NLP Model
• RoBERTa NLP Model

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Developed by Fakhrul Fauzi