News Bias Detector

This project is a machine learning-based solution to predict political bias (left-biased or right-biased) in textual content. The model is trained on labeled data using XGBoost. The frontend is built using Next.js, and a Streamlit app is maintained as a prototype.

Table of Contents

1. Overview
2. Features
3. Tech Stack
4. Workflow
5. Usage
6. Future Improvements
7. Contributing

Overview

In an era of diverse opinions and media influence, detecting political bias in textual content can provide insights into media and publication tendencies. This project utilizes a machine learning approach to categorize text as either left-biased or right-biased. By analyzing large volumes of labeled text data, the model aims to capture distinct features that differentiate these biases.

Features

• Text Preprocessing: Cleans and preprocesses raw text to optimize model training and prediction.
• Machine Learning Pipeline: Uses the XGBoost algorithm for classification.
• Bias Prediction: Predicts political bias (left or right) based on input text.
• Frontend Interface: User-friendly interface built with Next.js for making predictions directly from a web browser.
• Prototype Interface: Streamlit app for initial testing and prototyping.
• Backend Services: Flask backend for API management and integration.
• Database Management: Utilizes MongoDB for data storage alongside legacy JSON implementations.
• Workflow Management: Structured pipeline for reproducible and maintainable code.

Tech Stack

• Python: Core programming language.
• XGBoost: Model training and classification.
• Next.js: Frontend web framework for building the user interface.
• Streamlit: Prototype web application for initial testing.
• Flask: Backend framework for API development.
• MongoDB: NoSQL database for data storage.
• Scikit-Learn: Data preprocessing and evaluation utilities.
• Pandas, NumPy: Data handling and manipulation.
• Matplotlib, Seaborn: Visualization of data distribution and model performance.

Workflow

The project workflow is as follows:

1. Data Ingestion: Raw data collection and loading.
2. Data Transformation: Text cleaning, lemmatization, and vectorization.
3. Model Selection and Training: Training the model using XGBoost.
4. Evaluation: Model evaluation with metrics such as accuracy score.
5. Backend Development: Setting up Flask APIs for model interaction.
6. Frontend Development: Building the user interface with Next.js.
7. Database Integration: Managing data with MongoDB and maintaining legacy JSON implementations.

Clone the Repository

git clone https://github.com/aarshgupta24/L-R-News-Classifier.git
cd L-R-News-Classifier

Install Dependencies

Create a virtual environment and install required packages:

python -m venv env
source env/bin/activate # On Windows use `env\Scripts\activate`
pip install -r requirements.txt

Usage

Live Link 🔗 fontend 🔗 backend

for the Bias Detection Wensite

Launch Prototype Streamlit App: if you only want want bias detetion on the particular url you want

cd prototype
streamlit run app.py

Launch the webside on local host

terminal window 1

cd frontend
npm install
npm run dev

terminal window 2

cd backend
python main.py

Future Improvements

1. Support for Neutral Bias Detection: Add a "neutral" category to detect texts that don't lean towards any specific bias.
2. Sentiment Analysis Integration: Incorporate sentiment analysis to capture emotional tone alongside bias.
3. Improved NLP Techniques: Explore advanced techniques like BERT or RoBERTa for better feature extraction.

Contributing

Contributions are welcome! Please fork the repository, make your changes, and submit a pull request.