Comment by Xiyuan Gao

I’d say there’s no single F1 score showing that a sarcasm detector is ready. For evaluating readiness, I think a system-level criteria is more suitable … it’s about if the system is robust across different speaking styles, and how it handles uncertainty.