BLEU

GPTKB entity

Statements (35)
Predicate Object
gptkbp:instanceOf machine translation evaluation metric
gptkbp:basedOn n-gram precision
gptkbp:citation over 10,000
gptkbp:commonIn natural language processing
machine translation research
gptkbp:contrastsWith gptkb:ROUGE
gptkb:METEOR
gptkb:TER
gptkbp:criticizedFor not capturing semantic adequacy
favoring surface similarity
gptkbp:fullName gptkb:Bilingual_Evaluation_Understudy
gptkbp:higherScoreMeans better translation quality
https://www.w3.org/2000/01/rdf-schema#label BLEU
gptkbp:includes brevity penalty
gptkbp:input machine translation output
reference translations
gptkbp:introduced gptkb:Kishore_Papineni
gptkbp:introducedIn 2002
gptkbp:notRecommendedFor sentence-level evaluation
gptkbp:openSource gptkb:Moses
gptkb:NLTK
gptkb:SacreBLEU
gptkbp:output score between 0 and 1
gptkbp:publishedIn gptkb:ACL_2002
gptkbp:relatedTo automatic evaluation
corpus-level evaluation
gptkbp:usedFor evaluating machine translation quality
gptkbp:uses geometric mean
brevity penalty
modified n-gram precision
gptkbp:bfsParent gptkb:ROUGE
gptkb:Text_Summarization
gptkb:Data-to-Text_Generation
gptkb:WebNLG
gptkbp:bfsLayer 6