Arabic noisy text correction app

This project focuses on correcting noisy Arabic text using deep learning and transformer-based models to improve accuracy and readability. It supports manual text input, OCR, audio files, live audio, and text file uploads.

What I did: Built a custom transformer from scratch and fine-tuned pre-trained models such as AraT5, compared performance of models to optimize accuracy, and developed an interactive Streamlit dashboard to make the system accessible and user-friendly.