📌 Legal Question-Answering System using RAG

---

🧠 Overview

This project is an AI-powered system that answers user questions from large legal documents using Retrieval-Augmented Generation (RAG).

Instead of manually reading lengthy documents, users can simply ask questions, and the system retrieves relevant information and generates accurate answers.

---

🎯 Objectives

• Extract meaningful answers from legal documents
• Reduce time spent on manual document reading
• Provide accurate and context-based responses
• Demonstrate the use of NLP and LLMs in real-world applications

---

⚙️ System Workflow

PDF → Text → Chunking → Embeddings → Vector DB → Query → Retrieval → LLM → Answer

---

🧱 Project Structure

legal-rag-project/
│
├── data/
│   └── sample.pdf
│
├── src/
│   ├── data_processing.py
│   ├── embedding_db.py
│   ├── query_retrieval.py
│   ├── llm_module.py
│   └── main.py
│
├── requirements.txt
├── README.md
└── .env

---

👥 Team Members & Roles

Name	Role
Bharani	Query Processing & Retrieval
Pranay	Embeddings & Vector Database
Sai Teja	Data Processing (PDF → Chunks)
Parin	LLM Integration & Prompt Engineering

---

🧠 Key Concepts Used

• Natural Language Processing (NLP) – Understanding human language
• Embeddings – Converting text into numerical vectors
• Vector Database – Storing embeddings for similarity search
• Retrieval – Finding relevant information
• LLM (Large Language Model) – Generating human-like answers
• RAG (Retrieval-Augmented Generation) – Combining retrieval + generation

---

🛠️ Technologies & Libraries

• Python
• pdfplumber / PyPDF2
• sentence-transformers
• scikit-learn
• google-generativeai (Gemini API)
• (Optional) FAISS / ChromaDB

---

🚀 Features

• Handles large legal documents
• Retrieves relevant information efficiently
• Generates accurate answers using LLM
• Modular and scalable architecture
• Easy to extend into web applications

---

🧪 How It Works (Example)

Document contains:

Fraud is punishable with imprisonment up to 3 years.

User Query:

What is the punishment for fraud?

System Output:

Fraud is punishable with imprisonment up to 3 years.

---

⚠️ Challenges

• Understanding complex legal language
• Avoiding incorrect (hallucinated) answers
• Handling very large documents efficiently
• Maintaining high retrieval accuracy

---

🔮 Future Enhancements

• Use FAISS for faster vector search
• Add Streamlit-based UI
• Highlight answers in document
• Add multilingual support
• Build real-time chatbot interface

---

🔧 Installation & Setup

1. Clone the Repository

git clone https://github.com/Bharanikumar55/Legal-Question-Answer-System-using-RAG.git
cd Legal-Question-Answer-System-using-RAG

---

2. Install Dependencies

pip install -r requirements.txt

---

3. Add API Key

Create a .env file:

API_KEY=your_api_key_here

---

4. Run the Project

python src/main.py

---

📌 Usage

• Place your PDF file in the data/ folder
• Run the program
• Enter your question
• Get accurate answers instantly

---

🧠 Key Insight

This project demonstrates how combining retrieval mechanisms with language models can significantly improve answer accuracy compared to standalone LLMs.

---

📌 One-Line Summary

An AI system that extracts accurate answers from legal documents using Retrieval-Augmented Generation (RAG).

---

🙌 Acknowledgements

• Hugging Face
• Google Generative AI (Gemini)
• Open-source NLP community

---