assoudi-typica-ai/Darija-NLP-Lecture

# NLP for Darija — Content Moderation Companion Notebook Companion notebook to the lecture *"NLP for Darija: Text Classification with BERT-Style Models — A Case Study in Content Moderation for Moroccan Arabic"* — Hicham Assoudi, PhD (Al Akhawayn University, July 2026). ## What it does Reproduces the lecture's opening exercise end to end: 1. Judge 4 real Darija/MSA comments yourself — toxic or not? 2. Run them through three commercial moderation systems (OpenAI, Mistral, Claude) 3. Run the Darija-specific model `TypicaAI/DarijaToxicityDetector` 4. Compare — and see what general-purpose APIs miss in Moroccan Arabic ## Requirements - Google Colab (free tier is enough) - Three API keys stored in **Colab Secrets** (🔑 icon, left sidebar), with *Notebook access* enabled: - `OPENAI_API_KEY` - `MISTRAL_API_KEY` - `ANTHROPIC_API_KEY` The Darija model itself is open and needs no key. ## Contact **Hicham Assoudi** — Founder & Applied AI Researcher, Typica.ai · PhD (AI/NLP) *Typica.ai* — Independent applied research initiative 📧 assoudi@typica.ai · Linkedin . 🌐 typica.ai · 🤗 TypicaAI on Hugging Face