BLEU and ROUGE are automatic scores that judge machine-written text by how much its words and phrases overlap with human-written reference text.
In plain words
BLEU was built to grade machine translation: it checks how many short word sequences in the output also appear in good human translations, sentence by sentence, then averages across the whole set. ROUGE is a family of similar overlap scores used mostly for summaries. Both are cheap and fast, but they only count matching words; they do not judge grammar or whether the text is understandable.
A quick example
Score a machine translation against a human one. If both say "the cat sat on the mat", BLEU is high; if the machine writes "a feline was seated on the rug", BLEU drops even though the meaning is close.
Where you'll meet it
- Machine translation research.
- Summarisation and other text-generation benchmarks.
Related
Sources
- DocsMetric card for BLEU, Hugging Face Evaluate (GitHub) · read 29 Sept 2026
- DocsMetric card for ROUGE, Hugging Face Evaluate (GitHub) · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.