N03-1024 al. ( 2002 ) have proposed an automatic MT system evaluation technique (
N04-1022 AMTA , 2003 ) . We expect new automatic MT evaluation metrics to emerge
D13-1025 the translations produced by any automatic MT system still remain below than
N07-1005 system parameters based on the automatic MT evaluation measures . Acknowledgments
N06-1057 translation ( MT ) community for automatic MT evaluation . A problem with ROUGE
N06-1058 correlate with its utility for automatic MT evaluation . Our results suggest
N04-1036 translation quality is measured by the automatic MT evaluation metrics , such as
C88-2160 be aecomplisheA by any existing automatic MT system . The problem remains
N04-1022 translation in two scenarios . Given an automatic MT metric , we design a loss function
D09-1074 so as to directly optimize an automatic MT performance evaluation metric
D10-1090 evaluation . Inspired by the success of automatic MT evalua - tion , Lin ( 2004 )
D12-1090 systems . The early seminal work on automatic MT metrics ( e.g. , BLEU and NIST
N09-2006 riddled error surface computed by automatic MT evaluation metrics . We showed
N04-1022 ) , to the problem of building automatic MT systems tuned for specific metrics
D11-1035 significantly improve the quality of automatic MT compared to BLEU , as measured
I05-5003 translations . Fortunately , the automatic MT evaluation techniques commonly
J03-3003 , approaches other than fully automatic MT might provide interesting characteristics
D14-1020 Judgment A common means of assessing automatic MT evaluation metrics is Spearman
D13-1011 by short term improvements in automatic MT evaluation metrics such as BLEU
N07-1005 tion . However , a measure for automatic MT evaluation that strongly correlates
hide detail