A Systematic Comparison of Smoothing Techniques for Sentence-Level BLEU

摘要

BLEU is the de facto standard machine translation (MT) evaluation metric. However, because BLEU computes a geometric mean of n-gram precisions, it often correlates poorly with human judgment on the sentence-level. Therefore, several smoothing techniques have been proposed. This paper systematically compares 7 smoothing techniques for sentence-level BLEU. Three of them are first proposed in this paper, and they correlate better with human judgments on the sentence-level than other smoothing techniques. Moreover, we also compare the performance of using the 7 smoothing techniques in statistical machine translation tuning.