When do you think the machine translations of languages like Japanese and Chinese be on par with human translations?
Can they ever be? At least not until we have a significantdevelopment in AI. As there is a lot of judgement involved, these words are spoken for the first time in the world and so language is really creative, two words can different meaning in different contexts and it would change everything if gotten wrong. Since the way you arrange words to make sentences, maybe thatsentence has never been made before, or in that meaning, so you cant just literally assign values to every word, but the system would have to be able to see context and then judge the meaning of the word.
There is a lot of development for languages like French and Spanish but there is little to no progress in languages like Chinese and Japanese. How many years would it take to actually make texts like novels coherent to read?
When the time for a machine uprising halt commenced Jokes aside(not funny at all I know)probably won’t see that in this lifetime
I asked the same question on Quora, but all the experts there only seem to answer on the languages like Spanish but never talk about Asian languages.
not in near future for as a art but acceptable TL, a year is enough for same groups if programmers really wants
Depends, if you take the average fan translation with a terrible editor then you're about on par with MTL. And that's not taking into account several advantages from MTL. for instance it's usually more 'raw' without some TL taking liberties to change it without the reader knowing anything at all. It's more consistent as well. I found though that even more important than the translation (assuming that it's not a really high quality one) is easy access to the raws and being able to see what has been translated into what. Systranet (and LNMTL that use it along with some fancy UI on the site) does this, you hover over some of the words English or Chinese and the Chinese or English characters/words will be highlighted so you know what has been translated into what. Unless we're talking about really good translations then as far as I can tell what separate how "good" a translation is* has a lot more to do with the editing, i.e. English, than the actual translation. *or at least how good readers think/say it is What novels have you read and which ones have you tried with MTL (Also what MTL have you tried)? Depending on what you think is coherent then I'd say the answer is probably "already" p.s. btw could be interesting to have a look at that quora thread you mentioned
If theres A.I already, MTL is as good as humans. If not... oh well. let`s just assumed that MTL is a kid learning how to speak.
When a computer can keep track of context then translations will get better. At the moment it parses things sentence by sentence and doing it that way can shoot out translations that can be incorrect. For instance: That girl is on fire. We don't really know if the girl is literally on fire, burning and in flames... Or it meaning that she is doing something incredible well based on just this sentence. It needs additional context to figure it out... There are subtle things that have to be kept track of that also makes it difficult such as word connotations and other things. Authors not following proper grammar in their own languages or using phrases or idioms. It can get worse on made up words... I think we are still pretty far out to have context being able to be properly understood by AI... It's one thing to parse a sentence by sentence like IBMs watson in answering a single question... It's another to read a whole paragraph from one language and try to shoot out what it really means in another and get all of the meanings right... Even professional translators differ from each other, you'd be surprised how one professional translator can have something different from another... Some don't even get it right whatsoever...
Honestly, it's not as difficult as ppl make it out too be, but people are too dumb with their approaches. Still, even then, it'll be reliably doable before 2030 given enough investment. Definitely before 2040 with minimal investment.