📄 Document Q&A Assistant

An end-to-end NLP-based web application that allows users to upload documents (PDF/TXT) and ask questions based on their content through an interactive UI.

Live Demo

🔗 https://document-qa-assistant.onrender.com

Live Demo Note

The application is deployed on Render's free tier.
If inactive, the server may take 30–60 seconds to wake up on the first request.

🚀 Features

• 📤 Upload multiple documents (PDF, TXT)
• 📄 Automatic text extraction
• ✂️ Text chunking for efficient processing
• 🔍 Keyword-based search (information retrieval)
• 🤖 Answer extraction from relevant content
• 🌐 Interactive web interface (no Postman required)
• ⚡ AJAX-based upload and query (no page reload)

🧠 How It Works

1. Upload Documents Files are uploaded and stored locally.

2. Text Extraction Extracts content from PDF and TXT files using PyPDF2.

3. Chunking Splits large text into smaller chunks for better retrieval.

4. Search (Retrieval) Matches user query with chunks using keyword overlap.

5. Answer Extraction Selects the most relevant sentence as the final answer.

🛠️ Tech Stack

• Python
• Flask
• HTML / CSS
• JavaScript (Fetch API)
• PyPDF2
• Basic NLP (tokenization, keyword matching)

🔌 API Endpoints

1. Upload Documents

POST "/upload"

• Upload one or more PDF/TXT files
• Files are processed and stored locally

Request Type: "multipart/form-data"

Form Data:

• "files": one or multiple files

Response:

Uploaded: file1.pdf, file2.txt

2. Ask Question

POST "/ask"

• Takes a user query and returns an answer based on uploaded documents

Request Body (JSON):

{ "query": "What are access modifiers?" }

Response (JSON):

{ "answer": "Access modifiers are keywords...", "source": "java1st.pdf" }

⚠️ Limitations

• Uses keyword-based search (not semantic understanding)
• May return approximate results for vague queries
• Data stored in-memory (resets when server restarts)

🔮 Future Improvements

• Semantic search using embeddings
• Integration with LLMs (GPT / Gemini)
• Persistent storage (database / vector DB)
• Improved UI/UX

💡 Key Learning

Built a complete pipeline from document ingestion → text processing → retrieval → answer generation, similar to real-world RAG (Retrieval-Augmented Generation) systems.

👩‍💻 Author

Simran Duggal

AI/ML Engineer

⭐ If you found this useful

Give this repo a ⭐ on GitHub!

📸 Output Screenshot

Please refer to screenshots folder for output screenshots