AI Rookies

ROUGE — Recall-Oriented Understudy for Gisting Evaluation

Fact

A score that checks how much generated text overlaps a reference answer.

In Plain Words

ROUGE is like a teacher with a neon highlighter. It checks how many answer-key words your summary hit.

You meet it in text summary tests. It is cheap and easy, but it does not truly understand meaning.

Related Concepts

BLEU
BLEU checks if the generated words are precise. ROUGE checks if the reference content was covered.

Text Summary
ROUGE is often used for summaries. It measures overlap with a reference summary.

Benchmark contamination
If reference answers leak into training data, ROUGE scores can look too high.