METEOR
A machine translation evaluation metric that jointly considers precision, recall and word order, introducing synonym matching and stemming on top of BLEU and achieving higher correlation with human judgments.
Timeline (last 90 days)
Reference-based metrics (such as BLEU, ROUGE, and METEOR) score by computing n-gram overlap or semantic similarity, but they reward surface imitation and cannot determine whether the content is faithful to the input graph.
The system achieved a METEOR score of 0.3542 on 50 held-out cases from unseen centers.
The BLEU-4 and METEOR routines reproduced in the study achieve machine-precision-level agreement with the reference implementation.
All Facts (3)
Reference-based metrics (such as BLEU, ROUGE, and METEOR) score by computing n-gram overlap or semantic similarity, but they reward surface imitation and cannot determine whether the content is faithful to the input graph.
50%UnverifiedThe system achieved a METEOR score of 0.3542 on 50 held-out cases from unseen centers.
50%UnverifiedThe BLEU-4 and METEOR routines reproduced in the study achieve machine-precision-level agreement with the reference implementation.
50%