TruthGuard: AI-Powered Fake News Detector

Project Overview

TruthGuard is an AI-powered system designed to combat fake news through advanced AI and content moderation. The project leverages technologies like Llama 3 to distinguish between authentic and misleading information in real-time across digital platforms. This README focuses on the backend implementation of the project.

Key Features

• Real-time detection of fake news using AI algorithms
• Content moderation to ensure trustworthy information dissemination
• User-friendly interface for seamless interaction and feedback (frontend implementation)

Technology Stack

• Python 3.11
• Flask (Web Framework)
• Together AI (Llama 3 model)
• Pandas (Data manipulation)
• Scikit-learn (Machine Learning)
• LlamaIndex (Knowledge graph system)
• Milvus (Vector Database)

Setup and Installation

1. Clone the repository
2. Install required dependencies:

   pip install -r requirements.txt
   

3. Set up environment variables:
   • TOGETHER_API_KEY: Your Together AI API key

Usage

To run the Flask application locally:

python app.py

The server will start running on http://localhost:5000.

Testing the API

You can test the API locally using Postman or any API client. Additionally, the backend is deployed on Render and can be accessed at:

https://truthguard-ai-fake-news-detection-with.onrender.com

Local Testing with Postman:

1. Open Postman.
2. Create a new POST request to:

   http://localhost:5000/classify
   

3. Set the request body to JSON format with an "article" key containing the news text:

   {
     "article": "Your news article text here"
   }

4. Send the request and check the response.

Deployed API Testing:

1. Open Postman.
2. Create a new POST request to:

   https://truthguard-ai-fake-news-detection-with.onrender.com/classify
   

3. Set the request body to JSON format with an "article" key containing the news text:

   {
     "article": "Your news article text here"
   }

4. Send the request and check the response.

API Endpoints

• /classify (POST): Classify a news article as real or fake
  • Input: JSON object with an "article" key containing the news text
  • Output: JSON object with a "classification" key indicating "real" or "fake"

Model Training

The Llama 3 model was fine-tuned on the WELFake dataset using Together AI's platform. The fine-tuning script is available in fine_tune_llama3.py.

Future Enhancements

• Implement the KNN model and RAG pipeline as described in fake-news-detection-full-code.py
• Integrate Milvus vector database for efficient similarity search
• Expand the system to handle multiple languages
• Implement user feedback mechanism for continuous learning

Team

• Reema Memon - NLP Engineer (Team Lead)
• Sami Raza - AI Developer
• Letlhogonolo Fanampe - AI Specialist
• Wajahat Ali Hassan - Backend Developer
• Muhammad Qasim - Frontend Developer
• Muhammad Hassan - Full Stack Developer

Contributing

Contributions to improve TruthGuard are welcome. Please follow these steps:

1. Fork the repository
2. Create a new branch
3. Make your changes and commit them
4. Create a pull request

License

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

Contact

For more information, please contact:

• Letlhogonolo Fanampe
  • Email: Lfanampe@gmail.com
  • GitHub: djpapzin
  • LinkedIn: linkedin.com/in/djpapzin

Thank you for your interest in TruthGuard!