Automatic evaluation of machine translation quality
This note presents an overview of the problem of evaluating Machine Translation output. It reviews some of the main automatic metrics in use today. These automatic off-line evaluation metrics have served the Machine Translation community well over the past years and supported progress in statistical MT. However, in a number of contexts, there is a need to go beyond these existing metrics.
Presentation at the European Community
MTeval.pdf (76.38 kB)