Natural Laguage Question Answering System

This repository contains code, documents, references, and resources for the Major Project (Batch 2024).

Overview

Natural Laguage Question Answering System is a Flask web application that enables users to upload PDF files and ask questions related to their content. It leverages NLP techniques, particularly the fine-tuned Bidirectional Encoder Representations from Transformers (BERT) model, for accurate question answering.

Features

• Upload PDF files
• Ask questions related to uploaded PDF content
• Obtain answers extracted from the PDF using BERT
• Handle long documents with intelligent splitting and context preservation
• Display answers with relevant context from the PDF
• Easy-to-use web interface

Architecture

This client-server system comprises a Flask backend and a web frontend. Key components include:

• BERT Model: The fine-tuned BERT model converts questions and document content into semantic representations for accurate answer identification.
• Document Processing: The system employs the expand_split_sentences function to handle long documents exceeding BERT's token limit. It intelligently splits paragraphs while preserving context.
• Answer Extraction: Combining extractive and abstractive summarization techniques, the system provides concise and informative answers from relevant passages.
• Web Interface: The Flask backend serves a user-friendly interface for uploading PDFs, entering questions, and viewing extracted answers with context.

BERT Architecture (Optional)

For a deeper understanding of BERT architecture, you can refer to this explanatory resource.

System Design

The project adheres to a well-defined system design documented in the Major Project PPT file, including:

• Use Case Diagram: Illustrates interactions between users and the system (e.g., uploading PDFs, asking questions, viewing answers).
• Sequence Diagrams: Describes interactions between system components (e.g., file upload handling, question processing, response rendering).
• Flowchart: Provides a visual representation of program execution, including operations and decision points.
• ER Diagram (if applicable): Depicts data entities and their relationships within the system.

Results

Here are some screenshots showcasing the Document Question Answering System:

Home Screen

Upload PDF File

File Selected

Question Answering

Installation

1. Clone the repository:

git clone https://github.com/MohdSiddiq12/Major-Project

Note: Use code with caution.

2. Install dependencies:

pip install -r requirements.txt

Note: Use code with caution.

3. Download the pre-trained BERT model and tokenizer (replace with download instructions). Save them as bert_model.pkl and bert_tokenizer.pkl in the project directory.

4. Run the Flask application:

python app.py

Note: Use code with caution.

5. Access the web interface: http://localhost:5000 in your browser.

Contributing

Contributions are welcome! If you encounter issues or have suggestions, please open an issue or submit a pull request.

License

This project is licensed under the Apache License.