Research on Uyghur-Chinese Neural Machine Translation Based on the Transformer at Multistrategy Segmentation Granularity

Table 2

Test results.

Task	Model	Different granularity	BLEU %
Uyghur-Chinese	Self-attention-RNN	Syllable	12.68
		Marked syllable	47.98
		Words	45.61
		Syllable word fusion	51.74
	Transformer	Syllable	14.7
		Marked syllable	48.12
		Words	46.09
		Syllable word fusion	52.21
Chinese-Uyghur	Self-attention-RNN	Syllable	8.68
		Marked syllable	61.96
		Words	60.45
		Syllable word fusion	63.41
	Transformer	Syllable	11.19
		Marked syllable	63.01
		Words	62.81
		Syllable word fusion	63.88