This project collects human annotations for French memes across three risk dimensions: harmful, hateful, and propagandistic.
For each task, annotators review: - the meme image - OCR text, if available - a short description
They then assign:
- risk_presence
- is_harmful
- is_hateful
- is_propagandistic
- severity
- satire_humor
- context_required
- confidence
Please use the tutorial tab for the full annotation guideline and decision rules.
