Misinformation Detection with BERT + Flask + OCR

This project is a Fake News / Misinformation Detection Web App built with BERT Transformers, Flask, Google Search API, and OCR (Tesseract).

It allows users to: • Enter text manually or upload an image containing text. • The app extracts text (using Tesseract OCR) if an image is uploaded. • Detects whether the news/text is True or False using a fine-tuned BERT model. • If the text is classified as True, it fetches and displays 3 relevant verified links from Google.

⸻

🚀 Features • 🧠 BERT Transformer Model – Detects misinformation. • 🌐 Google Custom Search API – Provides verified links for true information. • 📸 OCR Support (Tesseract) – Extracts text automatically from uploaded images. • 🌍 Web Interface (Flask) – Simple and interactive UI.

⸻

🛠 Tech Stack • Python (Flask, PyTorch, Transformers) • BERT (distillBERT-misinformation-classifier) • Google API (Custom Search Engine) • OCR (Tesseract + Pillow) • Frontend: HTML, CSS, JavaScript

⸻

📂 Project Structure

📦 Misinformation-Detector ├── app.py # Flask backend ├── templates/ │ └── webpage.html # Frontend HTML ├── static/ │ ├── styles.css # CSS for styling │ └── script.js # JS for API requests ├── serious-energy-432707-a3-1c58b80fdb6f.json # Google API credentials └── README.md # Project documentation

⸻

⚙ Installation & Setup 1. Clone the repo

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

2.	Install dependencies

pip install flask torch transformers google-api-python-client google-auth pytesseract pillow

3.	Install Tesseract OCR

•	Download Tesseract for Windows
•	Place it in D:\Misinformation Classifier Project\tesseract.exe
•	Or update the path in app.py:

pytesseract.pytesseract.tesseract_cmd = r"D:\Misinformation Classifier Project\tesseract.exe"

4.	Run the Flask app

python app.py

5.	Open in browser

Go to 👉 http://127.0.0.1:5000/

⸻

🖼 Screenshots

🔹 Enter text or upload image 🔹 Get prediction (True/False) 🔹 See verified links for true news

⸻

🔮 Future Improvements • Deploy on Heroku / Render / AWS • Add multi-language OCR support • Improve model accuracy with more training • Store results in a database

⸻

📜 License

This project is open-source and free to use under the MIT License.