DocQA: Document Question Answering System

DocQA is an interactive web application built using Streamlit, designed to provide question-answering capabilities on uploaded documents. It utilizes pre-trained language models from Hugging Face's Transformers library to extract answers from PDF documents.

Details about the fine-tuned BigBird model can be found here

Features

• Model Selection: Allows to copy any Extractive QA model from Hugging Face link.
• Document Upload: Users can upload PDF documents to the system.
• Interactive Q&A: Users can ask questions and receive answers based on the content of the uploaded document.
• Highlighted Answers: The application highlights answers directly in the uploaded document for better context.

Demo 📽️

A demo of the system functionalities can be found here

Installation

1. Clone the Repository

   git clone https://github.com/najjarfred/DocQA.git
   cd DocQA

2. Install Dependencies

   Ensure you have Python 3.6+ installed, then run:

   pip install -r requirements.txt

   This will install Transfomers, Pinecone, Streamlit, PyPDF2, Pandas, and other necessary libraries.

Usage

To run the application:

streamlit run app.py

Navigate to the local URL provided by Streamlit, typically http://localhost:8501.

How It Works

• Upload a PDF: Start by uploading a PDF file.
• Select a Model: Choose a question-answering model from the sidebar.
• Ask Questions: Type your question into the input box.
• View Answers: The application processes the document and displays the answer, highlighting the relevant section in the document.

File Structure

• app.py: Main Streamlit application script.
• document_uploader.py: Handles document uploading and processing.
• qa_system.py: Contains the logic for the question-answering system using Hugging Face models.
• requirements.txt: Lists all the Python libraries that the project depends on.
• style.css: Contains custom CSS for styling the Streamlit app.

Contributing

Contributions to DocQA are welcome! Please read our Contributing Guide for details on how to contribute.

Acknowledgements

• BigBird for the pre-trained QA model.
• Streamlit for the web framework.
• Hugging Face's Transformers for pre-trained models.
• PyPDF2 for handling PDF files.