Question Just how good is google when Translating Japanese

Just how good is google when translating japanese now or in the future is it in the level of replacing human translators? I've started learning japanese for a couple of months now cuz i want to read japanese novels ,manga raw so is it too late for me or what im been doing is in vain?

 

so this is your text translated in jp and then in anglish again..... simple right? but....
When translating Japanese right now, or when translating in the future, how good is google is at the level to replace human translators? I have been learning Japanese for a couple of months. I'd like to read Japanese novels. Manga is not yet alive.

 

im pretty sure that unless a true AI is created or there is a big breakthrough in translating or quantum computing ( probably need both) then human translating will be a lot better even if a lot slower

 

What do you mean "too late"?
Well ithink google not really good with translating ,but its google they will find way

 

Not very. Ranges from somewhat readable for some series, to a complete jumbled mess.

In the future: maybe. In the foreseeable future: not at all.

Not at all. Learning the basics of the language will help you a Lot when trying to figure out the Google Translation: Google Translate sometimes gives you complete nonsense, but a quick look at the raws can be surprisingly helpful, even for very simple things like 'so there's an 'を' particle there, that was actually the object'.

Google translate is still helpful because you don't have to learn all the vocabulary, but knowing a bit of the language is extremely helpful.

I should know. Back when I was google-translating Korean stuff for myself, I learned some basic Korean grammar and that was super helpful.

 

its good to listen to it speak jap in case you read a character wrong i guess cuz some of those are just so hard to tell apart

 

Your doing isn't in vain. When you translate to another language, sometimes meaning behind is lost. Onii-chan , aniki

 

At some point they'll find an algorithm to divide the ambiguous double meaning words and auto-translate the rest, while letting translators or even the author choose what they meant from a checklist with potential words so it'll take 30 seconds a chapter.

 

this much is true as someone who doesn't know any language other than english and maybe spanish i use to try to google translate several Japanese novels in the past to get some spoilers. But the results were so bad that i almost broke my computer screen but since the recent update you can see there has been a lot of improvement in the translations.

 

What do you mean by foreseeable future? is it being close to that level in near future

 

It will take time as ai need to build up their understanding of words meaning and context. Did you know researcher have got 2 self learning ai to talk to each other they created their own language the research's didn't understand so they had to shut own the project.

I belive deep learning should let us translate documents on a fairy accurate rate within the next 10 years.

 

So guys what years is your estimation of it , for reaching the level of human translators?

 

Deep learning and neural networks in general are making very large leaps when it comes to understanding natural language problems such as translation. But it still has a bit to go (quite a large bit, but the prospects look hopeful). One large break through or method could easily make it jump by leaps and bounds, but as for a real time table its impossible to really know. The best thing to tell you is the current climate of the science behind it is that there are so many papers published that even people in the field can't read even 5% of them before more come out.

 

As soon as quantum computers become viable.

 

Well google created a quantum computers so does it count to being very near?

 

Just to give you a little taste, https://arxiv.org/list/cs/new , is a site that people post the rough drafts of yet to be peer reviewed papers. While this page lists all the new computer science related papers, this is over the span Thu 9 Nov 17 to Fri 10 Nov 17(two days), and has 169 entries... and a large number of these are about deep learning and neural networks (which are currently on the forefront of the translation effort)

 

It should be within a few decades, just look at how the first computers took entire rooms to computate simple math sums to where we are now with micro-chips doing complex computations. 20-30 years and it'd be viable to translate your book (assuming it's not a complex one with double meanings and wordplays) through google.

 

there are damnable people who ruining google translate with their suggestion, for example :
evil god => cthulu
when reading Dark Blood Age, qi/power i didn't check the raw => sex , what a great power => what a great sex!
When reading Sunrise in Another World, president house => Cheong Wa Dae , yes, it's Korean equivalent in romaji

 

Yes, when you digitize human intelligence you get.... human intelligence. The collective, as a whole, or rather the lack of it.

ATG: 'My chtulu seed has granted me such great sex!'

 

It means in the near future, you can forget about Google Translation replacing human translation. There have been a lot of improvements over the past decade, so it's pretty good now for English-French or English-Spanish and the likes, but Japanese-English? Eh... here's an example of a random google-translate from Japanese for example:

By the time Google Translate has become good enough to translate the above as well as a human translator, you'd probably have been able to learn enough Japanese to read it in Japanese by yourself.