http://daily.unitedlanguagegroup.com/stories/editorials/goog...
“…a multilingual NMT model trained with Portuguese>English and English>Spanish examples can generate reasonable translation for Portuguese>Spanish although it has not seen any data for that language pair. We show that the quality of zero shot language pairs can easily be improved with little additional data of the language pair in question,” the paper states.
I sometimes get disappointed when people suggest that machine translation is a substitute for actually learning the languages. It's true that we have limited time in our lives to study languages, and the machine translations can be a quicker way to get the gist, or possibly open some doors. But go much beyond that, like claims I sometimes see that you can wear AR goggles and suddenly communicate like a native, and you set unreasonable expectations. Or perhaps imagine a world where the way to read Shakespeare is to put it through a machine and it tells you, "statistically, Shakespeare meant to say this". Is that really reading it?
I imagine there are other languages around the world that share the same kind of similarities among themselves.
Same applies to other non-English languages, like Spanish and French, not sure how much this affects English variants as well.
Most of the time one gets a translation that isn't fully correct from the point of view of how the language is actually used in a specific country, sort of mixing American and British English on the same sentence.