salmabahdad/darija_translation_NLP

# 🇲🇦 Darija Translation Model (Darija → MSA / English) ## 📌 Project Overview This project aims to improve **machine translation for Moroccan Darija**, a low-resource and non-standardized dialect, by fine-tuning a **pre-trained multilingual transformer model (NLLB)** on a Darija-specific parallel dataset. General-purpose translation models are not optimized for Darija, which often results in **inconsistent and low-quality translations**. This project addresses that limitation by adapting the model to Darija data. --- ## 🎯 Objectives - Fine-tune a pre-trained multilingual model on Darija data - Improve translation **quality, fluency, and consistency** - Support: - Darija → Modern Standard Arabic (MSA) - Darija → English --- ## 🧠 Model Details - **Base Model**: NLLB (No Language Left Behind) - **Architecture**: Transformer (Encoder–Decoder) - **Training Type**: Supervised fine-tuning - **Framework**: Hugging Face Transformers ---