Chaima-KHENAFIF/algerian-darija-script-detector

Algerian Darija Script Detector

A fine-tuned xlm-roberta-base model that classifies text into three categories:

• Darija (Latin script) — Algerian Arabic written with Latin letters (e.g. wach rak)
• Darija (Arabic script) — Algerian Arabic written with Arabic letters (e.g. واش راك)
• Not Darija — any other language (French, English, Modern Standard Arabic, etc.)

Built to handle the heavy code-switching and script-mixing common in Algerian social media, messaging, and speech transcripts.

Usage

from transformers import pipeline

classifier = pipeline("text-classification", model="Chaima-KHENAFIF/algerian-darija-script-detector")
classifier("wach rak mlih?")

Training

Fine-tuned on a custom-labeled dataset of Algerian Darija and non-Darija text across the three classes above.

Results

Metric	Score
Accuracy	99.7%
F1 (macro)	99.7%
Precision (macro)	99.8%
Recall (macro)	99.6%

Links

• Full code, training pipeline, and dataset sample: GitHub repo