Audio Summarization Platform using Fine-tuned LLMs

This repository contains my master's thesis project from the National Polytechnic School of Oran, Algeria, focusing on exploring the fine-tuning process of Large Language Models (LLMs) for audio summarization.

Project Overview

The project aims to develop a web-based audio summarization platform by leveraging the power of fine-tuned LLMs, specifically the Whisper model. The platform provides an efficient way to generate summaries from audio content.

Repository Structure

audio-summarization/
├── code/
│   ├── notebooks/        # Fine-tuning notebooks for Whisper model
│   └── app/             # Streamlit web application
├── dataset/            # Training dataset for model fine-tuning
└── audio_sunerisation.pdf  # Thesis documentation

Features

• Fine-tuning implementation of Whisper model
• Streamlit-based web interface
• Audio processing and summarization pipeline
• Custom dataset integration

Getting Started

1. Clone the repository
2. Install dependencies
3. Run the Streamlit app:

cd code/app
streamlit run app.py

Documentation

For detailed information about the project, methodology, and findings, please refer to the thesis document (audio_sunerisation.pdf).

Author

Oussama MAHDJOUR Master's Student - Information Systems National Polytechnic School of Oran, Algeria