Do you have adblock enabled?
 
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Tech Crunch)   Google unveils new real-time universal translator   ( techcrunch.com) divider line
    More: Cool, Pixel Buds, Translation, Pixel Buds team, live language translation, total battery time, Language, bluetooth headphones  
•       •       •

2013 clicks; posted to Geek » on 04 Oct 2017 at 7:59 PM (10 weeks ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



54 Comments     (+0 »)
 
View Voting Results: Smartest and Funniest


Oldest | « | 1 | 2 | » | Newest | Show all

 
2017-10-04 06:53:01 PM  
img.fark.netView Full Size
 
2017-10-04 07:00:53 PM  
"Oh dear," says God, "I hadn't thought of that," and promptly vanishes in a puff of logic.
 
2017-10-04 07:25:32 PM  
WANT! That would be so useful (if it really works).
 
2017-10-04 07:32:26 PM  
4.bp.blogspot.comView Full Size
 
2017-10-04 08:07:24 PM  
I wish they had listed the languages.
 
2017-10-04 08:08:17 PM  
img.fark.netView Full Size

Yawn.
 
2017-10-04 08:13:42 PM  
img.fark.netView Full Size
 
2017-10-04 08:17:46 PM  

Bowen: [4.bp.blogspot.com image 850x652]


Quark is totally going to exploit these humans!
 
2017-10-04 08:29:54 PM  
A: it's not Google.
B: They suck. They're beta as hell and kinda crappy.
Instant Translation - Right in your Ear!
Youtube TayZG3aFubs
 
2017-10-04 08:30:38 PM  
to bad they used the worst ever design for the earbuds.
Basically the same shape as the AirPods, god I hate those.
 
2017-10-04 08:35:06 PM  
I was wrong. Wrong manufacturer.
 
2017-10-04 08:42:55 PM  
So, Babel fish?
 
2017-10-04 08:58:53 PM  
The Google Translate app already really does this. It just knows how to pipe one side to your ear buds.
 
2017-10-04 09:00:42 PM  

ginandbacon: WANT! That would be so useful (if it really works).


It's Google translate. Of course it doesn't work.
 
2017-10-04 09:00:45 PM  
Bet it doesnt do Glaswegian.

Burnistoun S1E1 - Voice Recognition Elevator - ELEVEN!
Youtube 5FFRoYhTJQQ


\ ordered one anyway
 
2017-10-04 09:06:25 PM  

Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.


I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.
 
2017-10-04 09:09:01 PM  

Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.


They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.
 
2017-10-04 09:41:20 PM  

Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.


Oh I am seriously going to have to check this all out. I gave up on the translate features a while ago but this sound awesome. What languages are you using it for?
 
2017-10-04 09:43:20 PM  

Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.


I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.
 
2017-10-04 09:52:46 PM  

itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.


IDIOM.  Idiom will fark even the best of translators.

It's better than nothing, I'll grant, but there will never be a way to replicate the human element.  Languages are too changeable over time.
 
2017-10-04 09:55:55 PM  

itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.


OK... That's very, very impressive... but it does have massive issues. I was playing with it just the other day to illustrate this to a friend. I typed "I have to poop" in Japanese and by removing the period at the end or changing the level of politeness, I got "I am out for poo," "I am out of poo," "she seems to have poop," "it seems like poo," "poo seems to come out," and "poultry comes out."
 
2017-10-04 10:01:04 PM  
tagha'
 
2017-10-04 10:03:43 PM  

itcamefromschenectady: I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.


Well, sure, but try that with Cloudmark or Firetongue.

/obscure?
//hexapodia as key insight
 
2017-10-04 10:07:06 PM  

Likwit: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

OK... That's very, very impressive... but it does have massive issues. I was playing with it just the other day to illustrate this to a friend. I typed "I have to poop" in Japanese and by removing the period at the end or changing the level of politeness, I got "I am out for poo," "I am out of poo," "she seems to have poop," "it seems like poo," "poo seems to come out," and "poultry comes out."


Well, doesn't Google Translate base its learning mainly on written stuff? So you can't expect it to be at its best with short sentences that are often spoken but rarely written...and the shorter a sentence is, the less context there is to help, even if the translator is the most adept there it.
 
2017-10-04 10:12:12 PM  

itcamefromschenectady: Likwit: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

OK... That's very, very impressive... but it does have massive issues. I was playing with it just the other day to illustrate this to a friend. I typed "I have to poop" in Japanese and by removing the period at the end or changing the level of politeness, I got "I am out for poo," "I am out of poo," "she seems to have poop," "it seems like poo," "poo seems to come out," and "poultry comes out."

Well, doesn't Google Translate base its learning mainly on written stuff? So you can't expect it to be at its best with short sentences that are often spoken but rarely written...and the shorter a sentence is, the less context there is to help, even if the translator is the most adept there it.


But we're talking about a device that's meant to be used in conversation. So if you go to a bar in Osaka and a girl you're hitting on says "be right back, I have to poop," it will say "I'm going for a moment, popping starts."
 
2017-10-04 10:24:14 PM  
let's see...go with darmok and jalad or airplane translator comment?
 
2017-10-04 10:44:47 PM  

jaggspb: let's see...go with darmok and jalad or airplane translator comment?


  I speak jive. Do you speak Klingon?
 
2017-10-04 10:45:51 PM  
Take the ridiculously high error rate when detecting speech and add it to the stupidly high error rate when translating. This is going to suck like an open window on the space shuttle.
 
2017-10-04 11:06:05 PM  

Russ1642: Take the ridiculously high error rate when detecting speech and add it to the stupidly high error rate when translating. This is going to suck like an open window on the space shuttle.


Not to belabor the point, but I think you are at least a year or two out of date on both counts. It won't be perfect, but both technologies are easily within the realm of being useable for this with a decent level of quality and consistency.
 
2017-10-04 11:09:56 PM  

Likwit: But we're talking about a device that's meant to be used in conversation. So if you go to a bar in Osaka and a girl you're hitting on says "be right back, I have to poop," it will say "I'm going for a moment, popping starts."


Popping Starts is my Accept cover band name.
 
2017-10-04 11:15:57 PM  

plewis: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

IDIOM.  Idiom will fark even the best of translators.

It's better than nothing, I'll grant, but there will never be a way to replicate the human element.  Languages are too changeable over time.


You need to get on the beam, gramps
 
2017-10-04 11:49:13 PM  

Delta1212: Russ1642: Take the ridiculously high error rate when detecting speech and add it to the stupidly high error rate when translating. This is going to suck like an open window on the space shuttle.

Not to belabor the point, but I think you are at least a year or two out of date on both counts. It won't be perfect, but both technologies are easily within the realm of being useable for this with a decent level of quality and consistency.


I am in fact dictating this, using Google speech recognition software right now. I didn't say for that, though. Oh need, and never used to insert punctuation that way. Sometimes it messes up a bit.

Also if I talk too fast can't keep up with me, just like real people. But since my thumbs are so horrendously horrible, it seems just like a better way to make post. Since I only work on my phone. the word work and that sentence was actually me saying fark, which I had to type because it won't accept a non-standard words. I also cannot say sebi. That was supposed to say subby.

Still, I am tempted to get a microphone on my desk at work install Google Text-to-Speech on my PC and see if it helps me when I need to write things. But I bet it's terrible at scientific jargon. Thalamus. Fmri. TMS. Transcranial magnetic stimulation. Neurodegenerative.

You know, now that I think about it, I might have use quite as much unusual Dragon East Beach I thought I did.

Oh, that was a good one.
 
2017-10-04 11:51:37 PM  

chawco: Delta1212: Russ1642: Take the ridiculously high error rate when detecting speech and add it to the stupidly high error rate when translating. This is going to suck like an open window on the space shuttle.

Not to belabor the point, but I think you are at least a year or two out of date on both counts. It won't be perfect, but both technologies are easily within the realm of being useable for this with a decent level of quality and consistency.

I am in fact dictating this, using Google speech recognition software right now. I didn't say for that, though. Oh need, and never used to insert punctuation that way. Sometimes it messes up a bit.

Also if I talk too fast can't keep up with me, just like real people. But since my thumbs are so horrendously horrible, it seems just like a better way to make post. Since I only work on my phone. the word work and that sentence was actually me saying fark, which I had to type because it won't accept a non-standard words. I also cannot say sebi. That was supposed to say subby.

Still, I am tempted to get a microphone on my desk at work install Google Text-to-Speech on my PC and see if it helps me when I need to write things. But I bet it's terrible at scientific jargon. Thalamus. Fmri. TMS. Transcranial magnetic stimulation. Neurodegenerative.

You know, now that I think about it, I might have use quite as much unusual Dragon East Beach I thought I did.

Oh, that was a good one.


I have reread my own post, this time actually reading what it said instead of just trying to follow along with the text as it types out while I'm speaking. Weird thing the text-to-speech does is occasionally change of word after its initial typing it out.

Having re read that post, I do not believe I will acquire a microphone and install text to speech on my computer in order to help me write scientific papers at work. I have enough time proofreading when I type with my fingers. There were some epic fails in the above post.
 
2017-10-05 12:20:30 AM  
The translations followed about 1-2 seconds after the people finished their sentences.

So not at all real time.
 
2017-10-05 12:30:34 AM  
img.fark.netView Full Size
 
2017-10-05 12:41:42 AM  

Daddy's Big Pink Man-Squirrel: The translations followed about 1-2 seconds after the people finished their sentences.

So not at all real time.


Nonsense. It real(ly took) time.
 
2017-10-05 12:45:07 AM  

Daddy's Big Pink Man-Squirrel: The translations followed about 1-2 seconds after the people finished their sentences.

So not at all real time.


True real time translation is effectively impossible because of variable sentence structure between languages and the importance of context to get the full meaning.

There are many sentences that you cannot reasonably start translating until you know the full sentence. Until and unless we get predictive AI that can tell what you are going to say before you say it, this is about as close to real time translation as we're going to get.
 
2017-10-05 12:52:08 AM  

itcamefromschenectady: Likwit: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

OK... That's very, very impressive... but it does have massive issues. I was playing with it just the other day to illustrate this to a friend. I typed "I have to poop" in Japanese and by removing the period at the end or changing the level of politeness, I got "I am out for poo," "I am out of poo," "she seems to have poop," "it seems like poo," "poo seems to come out," and "poultry comes out."

Well, doesn't Google Translate base its learning mainly on written stuff? So you can't expect it to be at its best with short sentences that are often spoken but rarely written...and the shorter a sentence is, the less context there is to help, even if the translator is the most adept there it.


Exactly... which is why a Babel-fish style translator like this will never work well if it's simply translating each phrase or sentence without any other context.   Especially between languages that have fundamentally different ways of thinking, in the way Japanese has very strict levels of status and politeness, yet the object and subject of a sentence are almost always left out and inferred.  Compared to languages that even conjugate verbs and adjectives differently depending on the gender/plurality of the subject or object.  Add Idioms to that and translators still have a long way to go before sounding anything like human, for speech.

Not saying it's not useful... but it's nothing like as smooth as the demo seemed to show.
 
2017-10-05 12:54:28 AM  
i.makeagif.comView Full Size
 
2017-10-05 12:55:07 AM  
I can't wait for the release of these things.  If they're anything remotely near what they advertise, people will take a little bit to adjust to strangers being able to understand when you swap to a foreign language for privacy.
 
2017-10-05 12:59:33 AM  

plewis: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

IDIOM.  Idiom will fark even the best of translators.

It's better than nothing, I'll grant, but there will never be a way to replicate the human element.  Languages are too changeable over time.

https://www.youtube.com/watch?v=mY9gVIcRkkI
 
2017-10-05 01:25:05 AM  
No they're not.
They can't get simple dictation right. When I try to turn words to text via android or google search, it's ~70% Acura at best. Even allowing for impossibly perfect translation, just to humor them, the algorithm doesn't even let me transcribe two reasonably good English, let alone the additional burden of translation other language after. Additionally, the problems of homophones and proper nouns witch Google completely barks up. Invisible gleaming.

If they can dictate English words into individual idea memes, and then recite the resulting words back, only then with this kind of thing become possible, with the resulting text. Until then, this is a pipe dream. Wurtzbach.

//I'm dictating this to my phone now so I'll be a kick out of these replies.
 
2017-10-05 02:42:05 AM  

lindseyp: itcamefromschenectady: Likwit: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

OK... That's very, very impressive... but it does have massive issues. I was playing with it just the other day to illustrate this to a friend. I typed "I have to poop" in Japanese and by removing the period at the end or changing the level of politeness, I got "I am out for poo," "I am out of poo," "she seems to have poop," "it seems like poo," "poo seems to come out," and "poultry comes out."

Well, doesn't Google Translate base its learning mainly on written stuff? So you can't expect it to be at its best with short sentences that are often spoken but rarely written...and the shorter a sentence is, the less context there is to help, even if the translator is the most adept there it.

Exactly... which is why a Babel-fish style translator like this will never work well if it's simply translating each phrase or sentence without any other context.   Especially between languages that have fundamentally different ways of thinking, in the way Japanese has very strict levels of status and politeness, yet the object and subject of a sentence are almost always left out and inferred.  Compared to languages that even conjugate verbs and adjectives differently depending on the gender/plurality of the subject or object.  Add Idioms to that and translators still have a long way to go before sounding anything like human, for speech.

Not saying it's not useful... but it's nothing like as smooth as the demo seemed to show.


Thisssss. So much this.
 
2017-10-05 03:46:05 AM  
So I will finally be able to figure out what the Arab cab driver finds so important that he needs to be on his phone the entire drive?
 
2017-10-05 07:12:46 AM  
Sokath, his eyes open!
 
2017-10-05 08:22:48 AM  

lindseyp: itcamefromschenectady: Likwit: itcamefromschenectady: Likwit: Delta1212: Likwit: ginandbacon: WANT! That would be so useful (if it really works).

It's Google translate. Of course it doesn't work.

I don't know where you've been for the last year or so, but ever since they switched to a machine learning base for their translations, they have gotten a lot better.

I played around with a google translate a lot in my spare time both before and after the conversion and the difference was very noticeable.

They have gotten a lot better, but it's still absolutely terrible. Spanish is passable, but any language with Chinese characters or Korean are absolutely dreadful.

I translated your post from English -> Spanish -> Chinese -> Korean -> English and got:

"They have had a lot of improvements, but they are still absolutely terrible. Spanish can pass, but Chinese or Korean is absolutely terrible."

I think you're being overly critical, when that chain of translations leads to something perfectly intelligible.

OK... That's very, very impressive... but it does have massive issues. I was playing with it just the other day to illustrate this to a friend. I typed "I have to poop" in Japanese and by removing the period at the end or changing the level of politeness, I got "I am out for poo," "I am out of poo," "she seems to have poop," "it seems like poo," "poo seems to come out," and "poultry comes out."

Well, doesn't Google Translate base its learning mainly on written stuff? So you can't expect it to be at its best with short sentences that are often spoken but rarely written...and the shorter a sentence is, the less context there is to help, even if the translator is the most adept there it.

Exactly... which is why a Babel-fish style translator like this will never work well if it's simply translating each phrase or sentence without any other context.   Especially between languages that have fundamentally different ways of thinking, in the way Japanese has very strict levels of status and politeness, yet the object and subject of a sentence are almost always left out and inferred.  Compared to languages that even conjugate verbs and adjectives differently depending on the gender/plurality of the subject or object.  Add Idioms to that and translators still have a long way to go before sounding anything like human, for speech.

Not saying it's not useful... but it's nothing like as smooth as the demo seemed to show.


Some of those may be actual problems for translation software because an accurate translation requires context that simply is not present in the actual text.

Idioms, on the other hand, are not nearly as difficult as you are making them out to be. Modern AI-driven translation works on the level of phrases and sentences, not individual words, which means that it is fully capable of picking out longer phrases as being units of meaning in their own right independent of the constituent words, which is all you need to translate idiom properly.

Out of curiosity, I plugged "It's raining cats and dogs" in Google translate just now and I got the appropriate German idiom as a response, rather than a literal translation of the above sentence (roughly more equivalent to "It's pouring buckets.")

Now, I'm not claiming that the software is already perfect at this, and I bet if I rather though a bunch more idiomatic phrases, I would find many that tripped it up. What I am saying, though, is that this is no longer a particularly insurmountable problem the way it was for previous generations of machine translation.

The real trouble is going to be, again as you also pointed out, implied information necessary to a translation that cannot be reasonably inferred from the words alone or information that must be conveyed in the translation but is not present in the original because of differences in how the languages express things and what meaning is obligatory to include, like register or gender or any number of other things, which is sort of a special case subset of the first issue.

I suspect that machine translation will get better at that than I reasonably expect it to, but it will never be as good as a human translator who is able to pick up outside context clues when making the translation in order to pull in information not accessible to the machine.

I suspect machines will eventually be better at translating a page of writing than a human translator given the same page with no other context, but it's always going to struggle to do in-person translation as well as a human unless it is given some other sensory inputs at the least.

/Changeable language is not something that I think will be an issue as much, since that would affect a human translator at least to the same degree and possibly even moreso.
 
2017-10-05 10:15:13 AM  

lindseyp: Exactly... which is why a Babel-fish style translator like this will never work well if it's simply translating each phrase or sentence without any other context.   Especially between languages that have fundamentally different ways of thinking, in the way Japanese has very strict levels of status and politeness, yet the object and subject of a sentence are almost always left out and inferred.  Compared to languages that even conjugate verbs and adjectives differently depending on the gender/plurality of the subject or object.  Add Idioms to that and translators still have a long way to go before sounding anything like human, for speech.

Not saying it's not useful... but it's nothing like as smooth as the demo seemed to show.


How would it do translating, "Buffalo buffalo buffalo Buffalo buffalo"?
 
2017-10-05 10:19:22 AM  
img.fark.netView Full Size
 
2017-10-05 10:20:20 AM  
call me when it's built into a e-book reader... and NOT as a an on-line service.
 
2017-10-05 10:28:21 AM  

Slypork: lindseyp: Exactly... which is why a Babel-fish style translator like this will never work well if it's simply translating each phrase or sentence without any other context.   Especially between languages that have fundamentally different ways of thinking, in the way Japanese has very strict levels of status and politeness, yet the object and subject of a sentence are almost always left out and inferred.  Compared to languages that even conjugate verbs and adjectives differently depending on the gender/plurality of the subject or object.  Add Idioms to that and translators still have a long way to go before sounding anything like human, for speech.

Not saying it's not useful... but it's nothing like as smooth as the demo seemed to show.

How would it do translating, "Buffalo buffalo buffalo Buffalo buffalo"?


About as well as a human who was unfamiliar with that sentence.
 
Displayed 50 of 54 comments


Oldest | « | 1 | 2 | » | Newest | Show all


View Voting Results: Smartest and Funniest

This thread is closed to new comments.

Continue Farking

On Twitter





Top Commented
Javascript is required to view headlines in widget.
  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.

Report