Thread: Machine translations & Idiomatic expressions

During the last months, I've been trying to break a myth I've had over the years: machine translations just do not work.
I have to admit that I'm seing very good work done by machines, but wonder how to work on idiomatic expressions, regardless of the languague you're managing. There are some very popular which work fine, but I'm always a bit worried about relying on them when it comes to this type of scenarios.
One I have in mind now is "The elephant in the room", is clearly not "El elefante en la habitación" how it can be seen as the outcome using google translator....so what do you do in cases like this?

Hi Malterio,


I´ve started using machine translations recently too and even though I found very good results I would not rely on them 100%.
In the case of idiomatic expressions you can always apply post editing to check if the figurative meaning is accurate.

Hi!

These days translators handle huge amounts of words in little time. Of course, any CAT tool is helpful.

That said, quality is non-negotiable. Human supervision is of the essence. Machine-production is not intelligent, and that shows easily.

Below I show some examples of the typical parallel product a good software for machine translation offers:

1) Focus the couple on coping with their results and supporting one another.
Enfoque a la pareja que enfrente sus resultados y se apoye uno al otro.

2) Inform couples that it may also be necessary to be tested.
Informe a las parejas que podría también ser necesario que se les realice la prueba.

In the first example, note the "forced" structure for "enfoque". In the second example, note the wrong but parallel colocation of "también".

Thanks Reminder! this is a great example to share here. Maybe it would be a good idea for all of us to share some DO's and DON'Ts of MT. Cheers.

I have been trying out a few machine translators over the past weeks and found that often they work quite nicely. But only for simple and short text. Idioms were one of the things that machine translators struggled with. When I found an idiom in the text, I usually just translated it by hand - if I was not familiar with the idiom in question, I obviously had to search for an explanation first. Thankfully there are some idiom dictionaries both online and offline that can be used for this. I am not sure whether idioms are something that can be completely tackled by machine translators, as idioms don't always have the same structure, wording etc. So it probably is not a trivial task to automatically detect them, I guess.

The main problems consist in recognizing an idiom and in distinguishing idiomatic from non-idiomatic usage, something that machine translation just cannot do.

MT (with baseline engines like Google) is improving, even when it comes to idioms. Try Google for two idioms, one almost the same for both languages (en and es), and the other slightly the same (based on character count) but with quite a different meaning and that would certainly sound literal.

1) a needle in a haystack, translates pretty well as "una aguja en un pajar." However, I admit being literal is totally acceptable in a case like this.

2) to kill two birds with one stone, is correctly machine-translated as "matar dos pájaros de un tiro," something that literally would sound totally awkward, no matter how similar it looks like.