Here are some standard translation bench-marks in machine translation task, the baseline score and the download links are shown below.
Dataset | Size | src->tgt(BLEU) | src<-tgt(BLEU) | Download |
---|---|---|---|---|
WMT14 EN-DE | 4.5M | 27.30 | 31.30 | Dataset |
WMT16 EN-RO | 610K | 33.70 | 34.05 | Dataset |
WMT17 EN-VI | 130K | 30.04 | - | Dataset |
WMT18 EN-ZH | 230K | 23.30 | - | Dataset |