Multimodal Document Analyzer 🔍

A Python tool that analyzes PDF documents using both visual and textual features to detect risks, anomalies, and important patterns. Unlike traditional text-only analysis, this tool "sees" documents like humans do.

🚀 What Makes This Special

Traditional PDF analysis: PDF → Extract Text → Analyze Text
This approach: PDF → Extract Text + Visual Features → Combined Analysis

The analyzer can detect:

• 📝 Missing or suspicious signatures
• 🔴 Visual tampering indicators
• 💰 Financial risks and monetary amounts
• ⚖️ Legal risk keywords and patterns
• 🎯 Layout inconsistencies

📋 Prerequisites

• Python 3.8+
• Windows users need Poppler for PDF to image conversion
• Tesseract OCR (optional, for enhanced text extraction)

🛠️ Installation

1. Clone this repository:

git clone https://github.com/chintanparekh2510/pdf-risk-analyzer.git
cd pdf-risk-analyzer

2. Install Python dependencies:

pip install -r requirements.txt

3. For Windows users:
   • Download Poppler from here
   • Extract and add the bin folder to your PATH

🎯 Quick Start

1. Place your PDF file in the project directory and name it test_document.pdf (or run python create_test_pdf.py to generate one)

2. Run the analyzer:

python document_analyzer_simple.py

3. Check the results:
   • Console output shows the formatted report
   • analysis_results.json contains detailed analysis data

📊 Example Output

╔══════════════════════════════════════════════════════════════╗
║          MULTIMODAL DOCUMENT ANALYSIS REPORT                 ║
╚══════════════════════════════════════════════════════════════╝

📄 Document: contract.pdf
📅 Analysis Date: 2024-01-19

🎯 Overall Risk Score: 72.5/100
   Text Risk: 65/100
   Visual Risk: 85/100

📊 Document Statistics:
   • Pages: 5
   • Words: 2,341
   • Risk Keywords: 13

💰 Monetary Amounts Found:
   • $5,000,000
   • $250,000
   • $1.5M

🖼️ Visual Features:
   • Signatures Detected: 0
   • Official Stamps: 3
   • Layout Consistency: 0.82

⚠️ HIGH RISK: Recommend legal review before signing
📝 No signatures detected - ensure proper signing

🔧 How It Works

1. Text Analysis: Extracts text and searches for risk keywords, monetary amounts, and concerning clauses
2. Visual Analysis: Processes document as images to detect signatures, stamps, and visual anomalies
3. Risk Scoring: Combines both analyses to generate a comprehensive risk score
4. Recommendations: Provides actionable insights based on findings

🎨 Customization

You can modify risk keywords and visual patterns in the MultimodalDocumentAnalyzer class:

self.risk_keywords = [
    'liability', 'penalty', 'breach', 'termination', 
    # Add your domain-specific keywords
]

self.visual_patterns = {
    'signature_region': {'min_area': 5000, 'aspect_ratio_range': (2, 6)},
    # Adjust detection parameters
}

📝 Use Cases

• Contract Review: Identify high-risk clauses before signing
• Document Verification: Detect potential tampering or forgery
• Compliance Checking: Ensure documents meet visual standards
• Due Diligence: Quick risk assessment of legal documents

⚠️ Limitations

• This is a demonstration tool, not a replacement for legal review
• Visual analysis accuracy depends on document quality
• Best results with standard business documents

🤝 Contributing

Feel free to submit issues and enhancement requests!

👨‍💻 Author

Created by @ChintanParekhAI

📄 License

MIT License - feel free to use in your projects!