BLEU is designed to approximate human judgement at a corpus level, and performs badly if used to evaluate the quality of individual sentences.
https://en.wikipedia.org/wiki/BLEU
To produce a score for the whole corpus the modified precision scores for the segments are combined using the geometric meanmultiplied by a brevity penalty to prevent very short candidates from receiving too high a score.
时间: 2024-10-16 11:13:06