Jenniyaa/darija-arabic-nlp-annotator

# Darija & Arabic NLP Annotator A lightweight, browser-based annotation tool for Arabic, Moroccan Darija, and French-Arabic code-switched text. Paste raw text, get structured NLP-ready annotations — no backend, no setup, no dependencies. # Darija & Arabic NLP Annotator A browser-based pre-annotation tool for Arabic, Moroccan Darija, and French-Arabic code-switched text. Paste raw text, get structured annotations — then review, correct, and export. This isn't a replacement for human annotators. It's an accelerator — it gets you 80% of the way there and flags the spots where you need to pay attention. **Live Demo →** · **Report a Bug** --- ## Why This Exists Most annotation workflows aren't built for Darija. Arabic gets partial support. Darija almost never. Darija mixed with French in the same sentence? Nothing handles that. But that's how Moroccans actually write online. "mzyan bzzaf had lfilm, tbarkallah 3la les acteurs" is a real sentence — Darija in Latin script, code-switching into French mid-sentence. This is the norm, not an edge case. This tool was built out of that frustration. It takes raw text and returns structured annotations with sentiment, language detection, intent, topic tags, and quality flags — formatted for direct use in any annotation platform or NLP pipeline. --- ## Features - **Multi-dialect support** — Arabic (MSA), Moroccan Darija, French, and mixed/code-switched text - **Franco-Arab handling** — Correctly processes Darija written in Latin script (3, 7, 9, etc.) - **Cautious Darija annotation** — System prompt tuned to default to lower confidence and flag slang, rather than wrongly assigning High confidence to ambiguous expressions - **Two annotation levels** — Basic (sentiment + language) or Detailed (sentiment, intent, topics, flags, notes) - **Batch processing** — Annotate one entry or hundreds at once, one per line - **Export options** — Copy to clipboard, download as JSON, or download as CSV - **No API key stored in code** — Users enter their own Groq API key in the UI - **Zero setup** — Single HTML file, runs in any browser, no server or build step required --- ## Limitations & Honest Notes This tool uses an LLM under the hood, and LLMs have real blind spo …