Skip to Main Content
Table 5: 
SacreBLEU of different models on WMT19 validation and test sets and perplexity per word of the language models on the English side of WMT19 validation set.
MethodModelUnpaired DataLM PPLTest17Test18Test19
Baseline transformer big – – 23.9 23.9 24.5 
 
This work Doc-reranker WMT 106.3 24.9 26.0 27.1 
Gigaword + WMT 63.8 25.5 26.3 27.1 
MethodModelUnpaired DataLM PPLTest17Test18Test19
Baseline transformer big – – 23.9 23.9 24.5 
 
This work Doc-reranker WMT 106.3 24.9 26.0 27.1 
Gigaword + WMT 63.8 25.5 26.3 27.1 
Close Modal

or Create an Account

Close Modal
Close Modal