Legal Document Importance Prediction

This project predicts the importance score of legal documents using
textual, structural, and semantic features.

Project Structure

• data/raw/ – Original datasets
• data/processed/ – Cleaned & feature-engineered data
• notebooks/ – Exploratory and experimental notebooks
• src/ – Production-ready pipeline scripts
• models/ – Trained CatBoost models
• outputs/ – Prediction outputs

Pipeline Steps

1. Data cleaning & normalization
2. Feature engineering (text, metadata, semantic)
3. Model training using CatBoost
4. Prediction

How to Run

python main.py

Model

• Algorithm: CatBoost Regressor
• Objective: RMSE
• Text features handled natively

Output

outputs/submission.csv contains predicted importance scores