Fake News Detector Web App

Welcome! This web application uses a hybrid approach developed from research using machine learning and large language models to help users determine the authenticity of news articles. The App also provides sentiment and topic analysis.

Features

• Article Classification: Determine if an article is real or fake using a trained logistic regression model.

• Fact-Checking: Leverage Meta's llama3-70b LLM model along with tools like DuckDuckGo and Wikipedia to fact-check claims from articles.

• Sentiment Analysis: Analyze the overall sentiment of the article (Positive, Neutral, or Negative).

• Topic Modeling: Identify key topics discussed in the article using Latent Dirichlet Allocation (LDA).

• Analytics Dashboard: View statistics on classified articles and trends over various time periods.

• Try it now at: https://fake-news-detector-web-app-v1.streamlit.app/

Installation if wanting to try locally:

1. Clone the repository:

   git clone https://github.com/yourusername/fake-news-detector.git
   cd fake-news-detector

2. Install the required packages:

   pip install -r requirements.txt

3. Run the Streamlit app:

   streamlit run HomePage.py

Usage

Once the app is running, you can:

1. Enter an Article or URL: Paste the full text of an article or a URL to an article. The app will analyze the content to classify it and provide additional insights.
2. View Results: See the classification (Real or Fake), sentiment analysis, key topics, and detailed fact-checking results.
3. Analytics Page: Explore analytics and trends about the articles processed by the app.

How It Works

Machine Learning Model

This Web App uses a logistic regression algorithm trained on approximately 76,000 news articles, achieving an accuracy of 0.95. The model uses TF-IDF vectorization for text preprocessing and classification.

Large Language Model (LLM)

For fact-checking, the app integrates Meta's llama3-70b LLM model. It extracts key claims from the article and uses external tools of DuckDuckGo and Wikipedia for verification.

Sentiment and Topic Analysis

• Sentiment Analysis: The sentiment of the article is determined using VADER sentiment analysis.
• Topic Modeling: Topics are identified using LDA, providing insights into the main subjects discussed in the article.

Technical Specifications

• Backend: Python, Streamlit
• Libraries: Langchain, scikit-learn, NLTK, requests, joblib, Streamlit-GSheets, and more.
• Data Source: The app uses a Google Sheets database for storing and retrieving classification results.

Contributing

Contributions are welcome! Please submit a pull request or open an issue to discuss any changes or improvements.

License

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

Acknowledgements

Special thanks to the developers of the libraries and tools used in this project.

---