Could this be considered a valid translation/back translation procedure if you translated a questionnaire from e.g. English to Chinese using one machine tranlsation engine and then back translated the obtained questionnare from Chinese to English using other machine translation engine; and resulted English version looks okay when compared with initial one?
Hello, there are controversial aspects on back-translating a machine-translated text by using (again) MT. MT back-translations may distort errors and make the process of correction more difficult. See
Nunes Vieira, Lucas, O’Hagan Minako and O’Sullivan, Carol (2021). Understanding the societal impacts of machine translation: a critical review of the literature on medical and legal use cases, Information, Communication & Society, 24:11, 1515-1532, DOI:10.1080/1369118X.2020.1776370
Hi Maxim Likhanov ,
Even though machine-translation is becoming more and more accurate, I think it wouldn't be a valid procedure, unless it was thoroughly tested for consistency. I mean, how could we be sure that an apparently good enough back translation wasn't but a random result? Would it have worked the same way for similar questionnaires? Would it always avoid pragmatic/cultural errors?
Translating a questionnaire, I believe, is meant not only to make it possible for the target users to answer it, but also for them to feel that the questionnaire matches their expectations in terms of how the questions are asked. The translation is therefore not only about the form (grammar/vocabulary), but also about style, and that is very hard (or perhaps still mostly impossible) to achieve with MT.
The short answer: yes. Because most translators today will use tools (more sophisticated ones than Googl Trans - like DeepL) to assist their translations.
The longer answer: to do so, one has to take into account the design and structure of the translation tool in question. Google does, for example, make use of publicly available parallel corpora (like those provided by the EU) which means that certain types of texts are almost fully human translated and G. is just the interface. The other key issue is that Google does NOT translate langauge-to-language (the kind of computing power needed for that would be prohibitively expensive). Instead, every translation is done via English (something that becomes apparent during back-translations). So, rather than say, Portugese-Russian, the translation chain is Portugese-English-Russian.
- Badges
- Science topic