The importance of this section lies in that the participating translators share their impression of using the customized NMT, a topic which rarely mentioned in previous research on the matter. If their feedback is consistent with the results of the evaluation mentioned in the previous section, it would be reasonable to say that the NMT can provide constructive assistance to translators. On the other hand, if the user experience of the
translators contradicts the result of the evaluation of the NMT, it would be a great opportunity to investigate reasons for the gap between the two.
The following table shows the two translators’ answers to each question. Please note that machine translation in Project 1 refers to NMT, while machine translation in Project 2 refers to Google Translate. However, the participating translators are not informed which MT text is shown in which project to not influence their feedback with any preconceptions or expectations of the automatic translation. The two participants answer every question with options provided by the questionnaire and do not fill in the blank field to provide additional thoughts.
Question Translator A Translator B
1 In Project 1 and 2, what is “the most”
helpful resource for improving translation work efficiency?
MT TM
2 Overall, is the machine translation provided in Project 1 coupled with TM helpful for improving translation work efficiency? helpful for improving translation work efficiency?
4 Which machine translation is more structure, which machine translation performs better?
Project 2 Project 2
7 In terms of punctuation, which machine translation performs better?
No differences No differences
8 Which machine translation can better compensate for the disadvantages of TM? following combinations can result in the best working efficiency?
Table 0.3 Translators’ answers to the questionnaire
For the first question, the two translators provide different answers. Translator A finds MT the most useful tool when choosing between TMs, MT and other internet resources, while Translator B prefers TMs. Their thoughts in this regard reflect their personal working
preferences: Translator A tends to make as little editing to the MT text as possible, while Translator B prioritizes compliance with the TM and is willing to make as many changes to the MT text as needed. The difference is especially obvious in Project 2, where the MT is generated by Google Translate. Please see the example below:
Google Translate: 這款光面平針織打底褲採用鬆緊腰身，拉鍊門襟和按扣設計，後口 袋設計。
Translator A: 這款亮面平織內搭褲採用鬆緊帶褲頭、拉鍊門襟和按扣設計，有後口 袋。
Translator B: 亮面平紋內搭褲，鬆緊帶腰圍，拉鍊門襟配按扣，有後口袋。
Next, the two translators agree on the second question. They both find the MT text provided in Project 1 has helped them spend less time on the translation, meaning the customized NMT does have a noticeably positive effect on their translation processes. As for the third question, Translator A still finds the MT text provided in Project 2 useful, but Translator B thinks the opposite. Translator A’s answers to these two questions are consistent with her feedback on the first one. Since Translator A acknowledges both of the MT text are effective in shortening translation time, it is reasonable that she decides MT the most useful tool among all resources mentioned. On the other hand, based on Translator B’s answer to the third question, it is presumable that in some aspects, editing the text generated by Google Translate is more time-consuming than the text from the NMT. The answers to the following questions might be able to provide further explanation of this phenomenon.
According to Translator A’s feedback on the fourth question, there is no obvious
Furthermore, considering her answers to the previous questions, Translator A most likely thinks both MTs are equally effective in reducing translation time and effort. In contrast, Translator B’s answers to Question 2, 3, and 4 show she acknowledges the customized NMT is more effective in increasing translation efficiency than Google Translate In the three questions that follow, the two translators basically share the same idea. The only disagreement is that Translator B thinks the NMT outperforms Google Translate in terms of diction while Translator A does not find either MT better. Interestingly, both translators prefer the performance of Google Translate in terms of grammar and sentence structure.
Lastly, it seems both types of MT can process punctuation appropriately. Based on these answers, it can be assumed Translator B spent their time and effort on editing terms in the text generated by Google Translate.
The two translators’ views diverge on the eighth question. Translator A finds Google Translate a better support tool when using TM to translate the projects, while Translator B believes the customized NMT provides more additional support. Although it is unsure the reason behind their difference in opinion, their answers to the next question may give some clues. Both translators observe the customized NMT’s translation is more similar to the text in the TM than that of Google Translate, even though they are not informed of the fact that the customized NMT and the TM are derived from the same corpus. Considering this, Translator A’s answer to the eight question could be based on the idea that Google Translate has successfully provided another version of an acceptable translation and thus can make a good complement to the TM. This idea also corresponds to the translator’s answers to the previous questions, showing she trusts the translation by Google Translate. In contrast, Translator B acknowledges the translation of the NMT is close to the text from the TM but
also provides extra help, implying the NMT has suggested better translations than the TM in some cases.
Last of all, the two translators’ answers to the final question provide defining evidence for identifying the effectiveness and necessity of either MT system in the experimental tasks.
Consistently, Translator A decides that Google Translate, TM and other internet resources are the most ideal tool combination for the translation projects. On the other hand, Translator B shows her preference for the combination of the customized NMT, TM and other internet resources. Although the translators have differing views in this regard, we can confirm that both MT systems are useful for the translators. Therefore, the follow-up discussion will focus on what really makes the customized NMT different from Google Translate.