I built this PDF chat tool about a year ago, before retrieval-augmented generation (RAG) became widely popular. Recently, I’ve updated it with BERT embeddings to go old school!

Key Features:

• Chat with PDF documents using natural language queries
• Powered by BERT embeddings for improved understanding
• Built without relying on external APIs

How It Works:

1. Upload a PDF
2. Ask questions about its content
3. Get relevant answers extracted from the document

Tech Stack:

• Python
• Streamlit
• PyPDF2
• BERT (Bidirectional Encoder Representations from Transformers)

This project demonstrates how classic NLP techniques can be combined with more recent models to create useful tools for document interaction.