BLEU Score Explainer
BLEU, short for bilingual evaluation understudy, is an automatic metric created for machine translation. It compares sequences of one or more tokens, called n-grams, in a candidate translation with n-grams found in one or more reference translations.
The score uses modified n-gram precision so repeated candidate words do not receive unlimited credit, and it applies a brevity penalty to discourage outputs that are too short. Corpus-level BLEU is generally more reliable than scoring one short sentence.
BLEU is fast and reproducible, which made it useful for comparing translation systems during development. However, valid translations can use different wording from the references, and high overlap can coexist with incorrect meaning, awkward grammar, or missed context.
The score depends on tokenization, case handling, reference set, n-gram order, smoothing, and implementation. Reports should document those choices and combine BLEU with human assessment and other metrics rather than treating one number as a complete judgment of quality.
BLEU is an automatic text-generation metric that compares n-gram overlap between candidate translations and reference translations.
Key terms include machine translation, candidate, reference, n-gram, modified precision, brevity penalty, corpus score, tokenization and smoothing.
The metric checks candidate n-grams against reference counts and combines precision across several n-gram sizes.
A brevity penalty reduces scores for outputs that gain precision by omitting too much content.
Reference wording is not the only valid wording, and surface overlap does not guarantee preserved meaning or fluent language.
Human review remains important for adequacy, tone, errors, and consequences that n-gram counts cannot capture.
Explore more "Explainers"
Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.
