Document Segmentation & Comparison Tool

Author: Arash Geramifard

Overview

This Streamlit application allows users to:

• Upload and segment documents (TXT, DOCX, PDF) into logical sections.
• Save and reuse segmentation templates.
• Upload filled documents and compare them against templates using OpenAI GPT-4 for semantic analysis.
• Get actionable feedback, remediation suggestions, and a match percentage for each section.
• Filter comparison results by status (e.g., Missing, Lacking Information, Sufficient, Other Issue).
• Download the comparison results as a JSON file.

Features

• Document Segmentation: Automatically splits documents into sections based on headers or numbering.
• Template Management: Save and reuse segmentation templates for future comparisons.
• AI-Powered Comparison: Uses GPT-4 to analyze filled documents, providing status, reasoning, remediation, and match percentage for each section.
• Visual Status Icons: Quickly identify section status with icons (✅ Sufficient, ❌ Missing, ⚠️ Lacking Information, ❓ Other Issue).
• Filtering: Filter results by one or more statuses for focused review.
• Downloadable Results: Export comparison results as a JSON file for record-keeping or further analysis.

How to Use

1. Install Requirements:

   • Install Python 3.8+
   • Install dependencies: pip install -r requirements.txt
   • Set your OpenAI API key in a .env file: OPENAI_API_KEY=your-key-here

2. Run the App:

   • Start the app with: streamlit run app.py

3. Segment a Document:

   • Choose "Upload new document for segmentation" and upload a TXT, DOCX, or PDF file.
   • Click "Segment Document" to split the document into sections.
   • Enter a filename and save the segmentation as a template.

4. Compare a Filled Document:

   • Choose "Use existing segmentation" and select a template.
   • Upload a filled document (same format as the template).
   • Click "Compare to Template" to analyze the filled document.
   • Review the results, filter by status, and download the JSON report if needed.

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Developed by Arash Geramifard with simplicity and best practices.