The reproduced blue score for En-Ro task is far away from the paper's report.

[hosseinbv]

Hi,
You have mentioned that you used mosesdecoder for computing bleu score, can you explain that? maybe this is the source of difference between my results and yours.

Thanks.

[jxhe]

Hi,

Mosesdecoder was the commonly used tool in machine translation research to compute the bleu score, and thus we use it in our paper.