News-Bias-In-India

Political Bias Detection In News Articles

This project uses traditional ML and BERT-based models to classify Indian English-language news as Left-wing, Right-wing, or Neutral. It incorporates sentiment analysis and manual labeling to fine-tune models for real-world applicability.

---

🚀 Features

• Accepts CSV or article text input
• CLI-based inference interface
• BERT fine-tuned for political alignment
• Stereo/anti-stereo sentiment analysis model
• Supports probability-based predictions with confidence warnings

---

🧠 Models Used

• bert_bias_tone_model – Trained on IndiBias_v1_sample.csv for sentiment tone (stereo/anti-stereo)
• bert_political_bias_model – Trained on 1000 manually labeled Indian articles (sourced via Mediastack API + manual annotation)

---

⚙️ Requirements

• Python 3.10+
• Libraries: transformers, torch, datasets, nltk, TextBlob, joblib, pandas

Install dependencies:

pip install -r requirements.txt