How good is Skype’s instant translation?
Thread poster: LegalTransform

LegalTransform  Identity Verified
United States
Local time: 03:01
Member (2002)
Spanish to English
+ ...
Mar 9, 2016

How good is Skype’s instant translation?

http://qz.com/526019/how-good-is-skypes-instant-translation-we-put-it-to-the-chinese-stress-test/


Direct link Reply with quote
 
Michele Fauble  Identity Verified
United States
Local time: 01:01
Member (2006)
Norwegian to English
+ ...
MT fallacy Mar 9, 2016

"First, the software has to recognize what we’re saying. It has to then figure out what that means, and convey the same meaning as best it can in another language."

There is no way that a computer can "figure out what that means".


Direct link Reply with quote
 

Philip Lees  Identity Verified
Greece
Local time: 10:01
Member (2008)
Greek to English
Wrong fallacy Mar 10, 2016

Michele Fauble wrote:

There is no way that a computer can "figure out what that means".



That's a bad choice of words in the article. There is also no way that a computer can "evaluate a board position in the game of Go", yet a computer has just won a game against one of the top players in the world.

Also, the concept of what something "means" is a purely subjective judgement, about which individual listeners may differ. We humans are no more able to comprehend what kind of subjective experience the computer is having when it translates something than vice versa.

Regardless of any considerations of computer "consciousness", I fully expect computer interpreting to be working at a practically satisfactory level within ten years at most.


Direct link Reply with quote
 

Rolf Keller
Germany
Local time: 09:01
English to German
Chess software does not translate Mar 10, 2016

That's a bad choice of words in the article. There is also no way that a computer can "evaluate a board position in the game of Go", yet a computer has just won a game against one of the top players in the world.


Right. But the sticking point is that with Go, Chess or whatever the target is exactly defined. In any given situation the software "simply" decides which one of, say, 99999 possible moves leads to a better position. So the software will be superior to a human if the software can "think" faster and/or evaluate more different moves than the human opponent. Actually, such software uses a large decision tree.

But how can a software decide whether translation A is more correct or better from any aspect than translation B? This would be a contradiction in terms. So, there is no decision tree at all, and no such game strategy works in the field of translation.

Actually, modern translation software products use a statistical approach. Google Translator works this way. Nasty spoken, this method simulates human translators, who rely on Linguee. The results are interesting, BUT:

- The methods rely on good & error-free corpora. Given the growing bulk of increasingly bad texts we are faced with in the modern world, where do these corpora come from in the future? Who will amend them daily, in a professional way?

- One of the problems is that texts don't contain all the info that is needed to "understand" it. Just look at the simple sentence "XXX is great". A software cannot know whether XXX is a singer, a cathedral, a landscape, a restaurant, a new smartphone, or ... Even if the software has access to a huge dictionary it just doesn't know that the speaker came back from Africa yesterday and thus is speaking about a landscape. Only the interlocutor knows this. But an adjective that fits for a singer might be totally wrong or even incomprehensible if attributed to a restaurant – esp if there are cultural differences.

A translator has the chance to grasp the meaning from the context, from the subsequent(!) text, from reference material, or from KuDoZ. A Skype interpreter lacks such chances. On the other hand, most informal conversations are very simple, compared to the texts most translators work on.

I think, a Skype interpreter for spoken communication would have to demand much language discipline from both partners. But at the beginning the "special" clientele, who likes playing around with such modern things, might get disappointed by such demands and thus the project might come to nothing.


Direct link Reply with quote
 

neilmac  Identity Verified
Spain
Local time: 09:01
Spanish to English
+ ...
Could do better Mar 10, 2016

It looks like it might have a better chance at success with more closely related languages, for example those using the Latin alphabet and stress- or syllable-timed rather than tonal. If nothing else, the Mandarin-English exchange could be good for a laugh.

Direct link Reply with quote
 


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

How good is Skype’s instant translation?

Advanced search







Anycount & Translation Office 3000
Translation Office 3000

Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.

More info »
CafeTran Espresso
You've never met a CAT tool this clever!

Translate faster & easier, using a sophisticated CAT tool built by a translator / developer. Accept jobs from clients who use SDL Trados, MemoQ, Wordfast & major CAT tools. Download and start using CafeTran Espresso -- for free

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search