🛡️ InfoSafe

InfoSafe is an NLP-powered Chrome extension that helps users assess the reliability of news articles by analyzing their content. Using a combination of keyword extraction, semantic similarity, and contradiction detection, InfoSafe delivers a percentage-based trust score that reflects how trustworthy a given article is. 🚀 Features

🔍 URL Analysis: Automatically fetches and analyzes the current tab's URL.

🧠 NLP Models: Uses HuggingFace Transformers for sentence embeddings and contradiction checks.

📊 Reliability Score: Combines semantic similarity, keyword matching, and contradiction detection to compute a trustworthiness percentage.

🌐 Trusted Sources: Compares news articles against established media like Reuters, AP News, BBC, and FactCheck.org.

🎨 Visual Feedback: Page background color changes from red to green based on reliability score.

🧠 How It Works

Content Extraction: The extension grabs the full text from the article currently open in the active tab.

Keyword Matching: Extracts key terms using NLTK to check for topic relevance.

Semantic Similarity: Compares the article against trusted sources using Sentence Transformers to detect factual alignment.

Contradiction Detection: Uses RoBERTa-large-MNLI to check for logical contradictions.

Score Calculation: Outputs a reliability percentage, factoring in:

    Content similarity (50%)

    Keyword alignment (20%)

    Contradiction absence (30%)
🛠️ Tech Stack

Frontend: HTML, CSS, JS

Backend: Python (Flask)

NLP Models:

    sentence-transformers/all-MiniLM-L6-v2

    roberta-large-mnli

Libraries:

    HuggingFace Transformers

    NLTK

    BeautifulSoup

    PyTorch

    Flask

    Pyngrok

🧪 Demo

💡 Start the backend server before launching the Chrome extension.

Clone this repo and run the backend Flask app:

python infosafe_backend.py

Start a tunnel (automatically done via pyngrok) and update the backend URL in:

    background.js

    manifest.json

Load the extension into Chrome:

    Go to chrome://extensions

    Enable Developer Mode

    Click "Load unpacked" and select the folder

🧑‍💻 Team

Johan – Project Lead & Backend NLP Developer

@simonknowsstuff – Frontend & Extension Integration

@AlvinGeorge-AG – UX Design & Data Flow
Nidhin Wilson - Developer