A Study on the Influence Mechanism of English Corpus on Translation Quality in Multilingual Website Translation

Figure 1.

Figure 2.

Figure 3.

Figure 4.

BLEU score of different NMT systems for NMT tasks

Compare	System	English-German	German-English	English-Vietnamese	Vietnamese-English	English-Russian	Russian-English
1	RNN	21.58	26.91	30.03	29.33	15.52	17.83
	Ours	24.49	28.32	30.85	30.54	15.99	19.48
2	Transformer	24.81	31.14	29.27	27.61	14.38	19.44
	Ours	24.88	32.89	30.04	29.15	15.21	20.14
3	mRASP	30.82	37.61	35.41	38.19	19.04	25.21
	Ours	31.85	38.02	36.47	38.78	20.15	25.44
4	mBART	29.81	37.74	35.13	37.64	19.18	25.53
	Ours	30.92	38.07	35.44	38.53	19.82	25.99

Comparison of completed translations

Project	Experimental group	Control group
Standard translation quantity	7047 words	7047 words
Total translations completed*	4358 words	5706 words
Percentage of translations completed/%	61.84	80.97
* χ2 = 742.65 ,P<0.05,there are significant differences.

Comparison of the number of people who completed all translation tasks

Group	Total number of people	Number of task completers	Percentage/%
Experimental group	40	3	7.5
Control group	40	21	52.5

Terminology translation statistics

Project	Experimental group	Control group
Correct quantity*	7047 words	7047 words
Untranslated and incorrect quantities	4358 words	5706 words
* χ2 = 2.84 ,P>0.05,there are no significant differences.

Comparison of translation accuracy of terms

Project	Experimental group	Control group
Number of terms of translation	363	502
Number of terms of the correct translation*	307	325
Accuracy rate of term translations /%	84.57	64.74
* χ2 = 38.15,P<0.05,there are significant differences.

Adding words to construct translated sentence data statistics

Project	Experimental group	Control group
Total number of translated sentences	204	241
Adjusted number of processed sentences*	124	113
* χ2 = 11.83,P<0.05,there are significant differences.