biko2020/Tkllm-darija

# Tkllm-darija — Moroccan Darija Data Factory > **A crowdsourced, production-grade platform for collecting, processing, and delivering high-quality Moroccan Arabic (Darija) datasets as a premium Data-as-a-Service (DaaS) for AI companies.** Tkllm-darija bridges the critical gap between Moroccan dialect data and modern AI systems. The platform delivers ethically sourced, culturally authentic, and continuously refreshed datasets to AI labs, enterprises, and governments building production-grade models for speech recognition, machine translation, virtual assistants, customer service automation, and more. Built with a focus on **scale**, **quality**, **cultural authenticity**, and **full regulatory compliance** (Morocco Law 09-08 / CNDP + GDPR-aligned practices). --- ## Table of Contents - Key Features - Platform Overview - Technical Architecture - Project Structure - Getting Started - Tech Stack - License - Contributing - Author --- ## ✨ Key Features ### For Contributors (Mobile-First App) - **Voice Recording** — Real-life scenario prompts: taxi orders, souk negotiations, medical consultations, code-switched conversations, and more - **AI-Assisted Transcription** — Automated first pass + human correction in Arabizi/Latin or Arabic script - **Rich Annotation Tasks** — Text labeling, translation, sentiment analysis, named-entity recognition, image/video description - **Built-in Darija AI Tutor** — Free chatbot for language practice and contributor onboarding - **Gamification & Rewards** — Streaks, leaderboards, and instant payouts via Orange Money, Inwi Money, or international transfer options - **Demographic Metadata** — Regional, age, gender, and accent data collected with explicit, granular consent ### For AI Companies (B2B Portal) - **Curated Datasets** — Speech-to-text pairs, parallel corpora, and fully annotated data packages - **Custom Collection Campaigns** — On-demand data collection tailored to specific domains or use cases - **Real-Time API Access** — …