A housing desk classifies emergency versus routine notes. Routine notes dominate, and a model that always predicts routine looks accurate while missing every emergency. Which metric surfaces the rare class?
Select an answer to reveal the explanation.
Short Explanation
Routine notes dominate, and always-predict-routine looks accurate while missing every emergency. On skewed text classification, prefer per-class or macro-F1 over headline accuracy. Exact match and perplexity on the common class hide the rare one.
Full Explanation
When one class dominates, headline accuracy can hide a model that never predicts the rare label. Per-class F1 or macro-F1 surfaces the emergency notes the housing desk actually cares about. Extractive exact match is a span metric, not a label metric. Perplexity on the common class does not score the rare one.