Contract Retrieval Tool

A tool for searching contract PDFs using natural language queries. Find termination clauses, renewal terms, payment details, and more using keyword-based search.

Features

• Extract text from PDF contracts once and save for fast searching
• Smart keyword expansion (e.g., "when does contract end" finds termination clauses)
• Interactive search mode or single-query mode
• Context preview around matches
• Page number references

Installation

Requirements

• Python 3.7+
• PyPDF2

Setup

# Install dependencies
pip install PyPDF2

# Or if using pip3
pip3 install PyPDF2

Usage

Step 1: Process your PDF contract

First, extract text from your PDF contract:

python3 process_contract.py your_contract.pdf

This creates a JSON file (e.g., your_contract_extracted.json) with the extracted text.

Example:

python3 process_contract.py DMS-2122-027CGroupDentalContract(Ameritas).pdf
# Creates: DMS-2122-027CGroupDentalContract(Ameritas)_extracted.json

Step 2: Search the contract

Interactive Mode

Run interactive search to ask multiple questions:

python3 search_contract.py your_contract_extracted.json

Then type your queries:

Search: when does this contract end
Search: what are the payment terms
Search: termination clause
Search: quit

Single Query Mode

Search with a single query:

python3 search_contract.py your_contract_extracted.json "termination clause"
python3 search_contract.py your_contract_extracted.json "when does contract end"
python3 search_contract.py your_contract_extracted.json "payment terms"

Example Queries

The tool understands common contract queries:

• Termination: "when does this contract end", "how to cancel", "termination clause"
• Renewal: "renewal terms", "automatic renewal", "how to renew"
• Payment: "payment terms", "fees", "billing", "cost"
• Term: "contract duration", "term length", "effective date"
• Liability: "liability", "indemnity", "damages"
• Confidentiality: "confidential information", "NDA terms"
• Insurance: "insurance requirements", "coverage"
• Notice: "notice requirements", "how to notify"

How It Works

Smart Keyword Expansion

When you search for "when does this contract end", the tool automatically expands this to search for:

• termination
• terminate
• cancel
• cancellation
• end contract

This increases the chance of finding relevant sections.

Relevance Ranking

Results are ranked by:

1. Number of keyword matches (pages with more matches appear first)
2. Context preview shows text around the first match

Project Structure

contract-retrieval/
├── process_contract.py          # Extract text from PDF
├── search_contract.py           # Search extracted text
├── extract_pdf.py               # Legacy: simple PDF extraction
├── contract_search.py           # Legacy: original search implementation
├── README.md                    # This file
├── your_contract.pdf            # Your PDF contract
└── your_contract_extracted.json # Extracted text (generated)

Future Enhancements (Option B)

Currently exploring semantic search with:

• sentence-transformers for better understanding of queries
• Vector embeddings for semantic similarity
• More natural language understanding

This requires additional dependencies and is planned for a future version.

Troubleshooting

"File not found" error

Make sure you run process_contract.py first to extract the PDF text.

No results found

• Try using different keywords
• Use more general terms (e.g., "payment" instead of "monthly payment")
• Check the PDF extracted correctly by opening the JSON file

PyPDF2 import error

pip3 install PyPDF2

Development Notes

Current Status (Option A - Keyword Search)

• ✅ PDF text extraction
• ✅ Keyword-based search
• ✅ Smart query expansion
• ✅ Interactive search mode
• ✅ Results ranking

Planned (Option B - Semantic Search)

• ⏳ sentence-transformers integration
• ⏳ Vector embeddings
• ⏳ Semantic similarity search
• ⏳ Question answering

License

This is a personal project for contract analysis.

Contributing

This is currently a personal tool. Feel free to fork and adapt for your needs.

---

Quick Start:

# 1. Extract PDF
python3 process_contract.py contract.pdf

# 2. Search
python3 search_contract.py contract_extracted.json

# 3. Ask questions
Search: when does this contract end?