📄 NLP-Based Contract Clause Classification & Entity Extraction

An end-to-end legal document analysis system that automatically classifies contract clauses and extracts legal entities using LegalBERT + NLP, wrapped in a clean Flask web application 🚀

---

✨ Features

• ✅ Upload PDF / TXT legal documents

• ✅ Automatic clause segmentation & classification

• ✅ Named Entity Recognition (NER) for legal entities

• ✅ Interactive UI

• ✅ Flask-based backend API

---

🧠 Powered By

• 🤖 LegalBERT (fine-tuned)
• 🧪 NLP pipelines for clause classification
• 🔍 Entity extraction (NER)
• 🐍 Python + Flask backend
• 🎨 HTML / CSS / Vanilla JavaScript frontend

---

🗂️ Project Structure

LegalBert/
│
├── backend/
│   ├── app.py                  # Flask backend
│   ├── models/                 # Trained LegalBERT model
│   └── utils/
│       ├── inference.py        # Model inference logic
│       └── pdf_extractor.py    # PDF text extraction
│
├── frontend/
│   ├── templates/
│   │   ├── index.html          # Home page
│   │   ├── analyze.html        # Analyze page
│   │   └── about.html
│   │
│   └── static/
│       ├── styles.css          # Global styles
│       ├── analyze.css         # Analyze page styles
│       └── analyze.js          # Frontend logic
│
├── uploads/                    # Temporary uploaded files
└── README.md

---

.pt file link

Download the .pt file from this link

---

🖥️ Frontend Highlights

📤 File Upload

• Drag & drop or Choose File
• Supports .pdf and .txt

📊 Results Summary

• Total clauses
• Clauses with labels
• Total entities
• High-importance clauses

📑 Clause Viewer

• Shows Top 10 clauses by default
• 🔘 View More → see all classified clauses
• 🔁 Toggle back to Top 10

🏷️ Entity Display

• Entity label + text

---

🚀 How to Run

1. Create virtual environment

   python3 -m venv .venv
   source .venv/bin/activate   # Windows: .venv\Scripts\activate

2. Install dependencies

   pip install -r requirements.txt

3. Add the classifier checkpoint

   Place the trained model at:

   backend/models/best_legal_classifier.pt
   

   This file is large and should be kept as a local artifact, not committed as a normal Git file.

4. Start the server

   cd backend
   python app.py

5. Open the app

   Visit http://127.0.0.1:5500.

Entity extraction uses Ollama at http://localhost:11434 with phi3:mini. If Ollama is not running, clause classification still works and entity extraction falls back to an empty list.

---

🎯 Use Cases

• 📜 Contract review
• ⚖️ Legal compliance analysis
• 📚 Academic legal NLP research
• 🏢 Enterprise contract intelligence

---

Contributors

• Aastha
• Niveditha

---

Demo Images