Your example of “judicieux” being translated as “wise” seems good to me; “judicious” and “wise” are very similar in meaning in English, and “wise” is much more common, and therefore the most natural translation for most situations - checking a couple of dictionaries, both “judicious” and “wise” are given as translations of the same meaning of “judicieux”, and the example sentences given exclusively use “wise”. The word “judicious” is also quite formal in style and not common in everyday speech in English, even to the extent that I suspect many English speakers may not be 100% sure what it means. I can’t speak for the Russian translations specifically, but the English back-translations you’ve given seem fine to me for those reasons. I’ll check with our Russian linguist for a better understanding of the Russian translations and whether there is a better alternative for them.
If you have any other examples of translation issues, please flag them via [this link] so that we can have the relevant linguists take a look. (Submit a translation error)
One thing to remember is that any translation of a simple word and many phrases on Memrise is usually lacking context, and so the way we go about translating those words in the absence of that context is to assess two things about the target item:
- What is the most ‘obvious’/common usage, i.e., what springs immediately to mind when most native speakers hear the word/phrase?
- What translation would best capture either this most common usage, or would fit the widest range of contexts?
When a dictionary gives you a translation of a word, it usually lists multiple options for this exact reason. However, when thinking about context, register, the best way to convey meaning to a learner, and the best option for many of our test types within Memrise (for example, you don’t want huge translations that take up half the screen in Speed Review), we may choose to opt for a more natural, less ‘standard’ translation, or vice versa depending on the item at hand.
As for our process for reviewing and correcting mistakes, we have a spreadsheet where we log issues that have been raised. Our internal linguists then review this once a month, involving external linguists for the languages we don’t cover internally. Since many translation issues are subjective for the reasons mentioned above, we therefore generally need some discussion between two linguists who are native speakers of the target and source language to nail the nuances of the phrase and decide how to proceed. Of course, whenever we are made aware of any errors issues outside of this cycle that we can fix right away, we always do so.
Unfortunately, we can’t always reply to everyone who has flagged mistakes to explain whether or how they have been addressed due to time constraints. Responding can often open up a long discussion about the nuances of translation, language usage and meaning, and why translations aren’t always clear-cut, black and white decisions.
I hope this helps you understand a little better how we approach translations and correcting errors