Back to LanguageTool Homepage - Privacy - Imprint

[req] English feedback on non-English grammar issues


(Lodewijk Arie van Brienen) #1

I've recently restarted work on a story where a part of it takes place in an Italian restaurant.
Part of the dialogue in this section is in machine-translated Italian, but all of the grammar warnings I get from it are completely unreadable to me.
So, would it be possible to add an option to A. show all the warnings in only English? or B. English along with the actual language?


(Daniel Naber) #2

That's not so easy, as it would mean that we have to translate many, many thousand messages. Also, if you don't speak a language you shouldn't blindly listen to LT anyway, i.e. some messages will need to be ignored.


(Lodewijk Arie van Brienen) #3

A. I understand that this will take a LOOOOOOOONG time. (If this were to be done, best to first focus on the so-called dominant languages.)

B. What I'm hoping for is 'just enough information' to cross-check with a few language sites for the 'simple' question of: Does the warning make sense? (even in English LT talks out of its ass every now and then, and with English I'd at least know where to look)

C (As a last resort) could you perhaps add an option to disable LT for specific languages?


(Martin Bodin) #4

Maybe I am misunderstanding your issue, but I have the feeling that you consider that if LT does not detect any mistakes in a machine-translated sentence, then it is a good translation. Well, one of the greatest critic of Google Translate (for instance), is that it often generates correct sentences, but with a meaning totally different from the inital sentence. For instance, from « La muson manĝas la kato. » (meaning “The cat eats the mouse.” in Esperanto), I get “The cat eats the cat.” as a translation (source: https://translate.google.com/?source=osdd#auto/en/la%20muson%20man%C4%9Das%20la%20kato ). The machine-translated sentence is correct (although strange), but does not mean the same than the initial sentence.

My conclusion is that the A and B options are dangerous: however good the results of LT are, it can not be fully trusted, and a speaker of the language should check the sentence. It is thus a good hypothesis (I think) to consider that the users of LT speak the checked language. Option C seems an interesting feature to me.

Best,
Martin.


(Lodewijk Arie van Brienen) #5

No, at the moment I only know that LT considers something wrong, but I prefer to get something readable so I can get second opinions from dictionary-sites. (I know from personal experience that manual-translation, while more time consuming, is much more reliable than machine-translation)

and when I do use machine translations I always cross-check with the back-translation.
(EG: I once tried to translate "radical" (measures) to Japanese, all machine translations at the time came with the back-translation of "radical character" (a base concept of Japanese and Chinese writing), I then looked it up in a dictionary-site and found only one translation that back-translated as "extreme")


(Jung Xiaohui) #6

That’s right, machine translation isn’t quite dependable. I’ve tried with GNMT the top player in the machine translation field (for English<>Chinese language pair) and the result isn’t ideal. For important stuff you still have to trust in human translations. :grinning: