A Joint Back-Translation and Transfer Learning Method for Low-Resource Neural Machine Translation

Table 1

The statistics of datasets.

Corpus	Dataset	Sentences (M)	Test set
English-Chinese	Union parallel data	15	Newtest2017
Turkish-English	WMT16 parallel data	0.2	Newtest2016
Uygur-Chinese	CWMT2017 parallel data	0.35	CWMT-2017
English	Monolingual data	0.75	N/A
Chinese	Monolingual data (organized)	0.9	N/A