Darija-Translator

A Nano-Transformer for Tunisian Darija-to-English translation, built from scratch. Pre-trained on Moroccan Darija data (RTX 3050 Laptop), then fine-tuned on hand-crafted Tunisian pairs (RTX 4070 Desktop).

Model Details

• Architecture: Encoder-Decoder Transformer
• Parameters: ~15.6M
• Tokenizer: SentencePiece BPE (16,000 tokens, Arabizi-aware)
• Training data: 35,977 cleaned Moroccan Darija pairs
• Fine-tuning data: 500 hand-crafted Tunisian pairs across 50 categories
• Train/val/test split: stratified 8/1/1 per category, honest evaluation
• BLEU score: 3.89 (v1 baseline, June 2026, test set never trained on)
• Best checkpoint: epoch 7/20 (by validation loss)

Limitations

500 Tunisian pairs is a small foundation. Simple sentences show recognizable structure; complex cultural expressions are still limited. The dataset is actively growing toward 1,000 pairs and every retrain will be documented with a new BLEU score.

Author

Dhia Azizi (@dhiadev-tn) GitHub: https://github.com/Dhiadev-tn/darija-translator Dataset: https://huggingface.co/datasets/Dhiadev-tn/tunisian-darija-english