Microsoft gives up on machine translation
| |
"Offering machine translation and let user’s offer a better translation (this way they improve their TMs)." I saw that, Gaston! How dare they! I would NEVER do such a thing for them, let them improve their TMs alone! Once I needed a simple letter translated into Spanish, a language that I do not master (although some people might thing, Portuguese - that I speak - and Spanish are NOT the same) and I used the Google Translation. Ops! What a blunder! The translation was so poor that I ... See more "Offering machine translation and let user’s offer a better translation (this way they improve their TMs)." I saw that, Gaston! How dare they! I would NEVER do such a thing for them, let them improve their TMs alone! Once I needed a simple letter translated into Spanish, a language that I do not master (although some people might thing, Portuguese - that I speak - and Spanish are NOT the same) and I used the Google Translation. Ops! What a blunder! The translation was so poor that I had to ask a human for correction. And, IMHO, if I'll have to ask a human to check a machine's work, this means the machine is not ready yet. And it will never be, again IMHO, because translation is not a repetitive task - it is a complex, thinking and intuitive task. Machines are only good with repetitive tasks. When someone invents a machine that can think, then most probably we'll have translator machines... ▲ Collapse | | |
"Microsoft gives up on machine translation" or maybe they'll just leave it to Google... | | | l Gaston l Local time: 06:45 Afrikaan menyang Ibrani Kuno + ...
Wow! I would say "Impossible is nothing", just give the proper time to machines...
Btw, Microsoft's translation widget looks similar to what Google has been doing for years: Offering machine translation and let user's offer a better translation (this way they improve their TMs).
I can see machines getting better on translation, I wouldn't be surprised if in the near future I can read any website in the world in my native language. | | | l Gaston l Local time: 06:45 Afrikaan menyang Ibrani Kuno + ...
I belive that, in the same way than humans, machines need to learn the language. And "they" are reading books while we speak. Sounds scary (and funny) some times, but I'm really anxious to see the results of "educated" machines. I belive Google is doing a great job currently. | |
|
|
Jeff Whittaker United States Local time: 05:45 Spanyol menyang Inggris + ...
The problem with the Google translate engine is that it only SEEMS to be doing a great job because it does not translate each word, but tries to find statistical matches for phrases. In other words, more of what Google translates will sound better. However, the problem is that sometimes the translation will sound great, but be complete wrong because the computer has simply made a "guess" at what the translation could be, rather than what it actually says. To liken this to human translation, if w... See more The problem with the Google translate engine is that it only SEEMS to be doing a great job because it does not translate each word, but tries to find statistical matches for phrases. In other words, more of what Google translates will sound better. However, the problem is that sometimes the translation will sound great, but be complete wrong because the computer has simply made a "guess" at what the translation could be, rather than what it actually says. To liken this to human translation, if would be as if the human just made something up. ▲ Collapse | | | Theo Bernards (X) France Local time: 11:45 Inggris menyang Walanda + ...
I agree with Jeff's comment here: Google seems to ignore and omit whatever it cannot translate. The grammar structure in my native language (Dutch) is fairly complicated and the result of Google translation of any long text is at best an approach of what the source text says. And to top things off with getting all that info for free by asking for a better translation by users, Google has created a Translators Toolkit as well, which will of course also work towards improving their TM. | | |
The scenario Jeff suggests is really quite scary. The meaning is in fact more important in most cases than idiomatic prose.
Half of translation is understanding the source correctly, and that is what machines cannot do. Even the best TM can come up with perfect matches that are simply wrong for the context. The human mind is so used to working in context that it often comes as a surprise that any confusion is possible. But the machine has no way of knowing - statistics alone are not... See more The scenario Jeff suggests is really quite scary. The meaning is in fact more important in most cases than idiomatic prose.
Half of translation is understanding the source correctly, and that is what machines cannot do. Even the best TM can come up with perfect matches that are simply wrong for the context. The human mind is so used to working in context that it often comes as a surprise that any confusion is possible. But the machine has no way of knowing - statistics alone are not nearly reliable enough.
This is blasphemy in some people’s ears, but in many situations a correct, but slightly unidiomatic rendering of the meaning is a better translation than elegant prose that does not accurately convey the meaning of the source. That is what the developers need to concentrate on.
I had a lovely case recently of people having to wear shorts to access a building. Strange dress code for Swedish businessmen, but it was absolutely mandatory for security… It finally dawned on me that the source undoubtedly said they had to have (magnetic ID) cards  ▲ Collapse | | | jaymin Canada Local time: 05:45 Jerman menyang Korea + ...
Nope!!!, eventually the machine translation will take over this industry. | |
|
|
I completely agree with Jeff: "The problem with the Google translate engine is that it only SEEMS to be doing a great job because it does not translate each word, but tries to find statistical matches for phrases. In other words, more of what Google translates will sound better. However, the problem is that sometimes the translation will sound great, but be complete wrong because the computer has simply made a “guess” at what the translation could be, rather than what it actually says. To li... See more I completely agree with Jeff: "The problem with the Google translate engine is that it only SEEMS to be doing a great job because it does not translate each word, but tries to find statistical matches for phrases. In other words, more of what Google translates will sound better. However, the problem is that sometimes the translation will sound great, but be complete wrong because the computer has simply made a “guess” at what the translation could be, rather than what it actually says. To liken this to human translation, if would be as if the human just made something up." ▲ Collapse | | |
Computers are the best at for complex and repetitive operations. Translation has nothing to do with that because it captures the real meaning of words and the context, which is infiniteley more complex than just putting words next to each other. For a computer to do as good as a human translator it would have to learn and understand all human history since the beginning of times. How can a machine capture the human spirit and emotions? And even if one day machines could capture the essence of tr... See more Computers are the best at for complex and repetitive operations. Translation has nothing to do with that because it captures the real meaning of words and the context, which is infiniteley more complex than just putting words next to each other. For a computer to do as good as a human translator it would have to learn and understand all human history since the beginning of times. How can a machine capture the human spirit and emotions? And even if one day machines could capture the essence of translation, it would certainly not be able to translate poems....Technical translations yes..Litterary work no because non technical documents by nature have different meanings according to the situation and the context.How would a machine know how a reader feels when reading Hamlet or another work like this? Translating is feeling (it's an art)..Machines don't feel, therefore will reach a limit to their potential of translation....Well except if the Terminator comes back to finish up the work ▲ Collapse | | | Igor Popov Russian Federation Local time: 12:45 Inggris menyang Rusia + ...
I agree with Alexandre that translating is feeling (it’s an art). 20 years ago I happened to meet a pioneer of translation memory or in other words a passionate advocate of what we call TM today. They were trying to develop a machine for translation of simple technical texts and even succeeded in doing so in a sense. However, they quit due to lack of financing.
With all respect to the guys who have been developing TM tools and the progress they have been making, I believe that at the m... See more I agree with Alexandre that translating is feeling (it’s an art). 20 years ago I happened to meet a pioneer of translation memory or in other words a passionate advocate of what we call TM today. They were trying to develop a machine for translation of simple technical texts and even succeeded in doing so in a sense. However, they quit due to lack of financing.
With all respect to the guys who have been developing TM tools and the progress they have been making, I believe that at the moment TM is a matter of business rather than a real assistance for translators. ▲ Collapse | | | John Fossey Canada Local time: 05:45 Member (2008) Prancis menyang Inggris + ...
Just as with human translators, I believe computer's will only gain native language proficiency when they have become natives of the language itself... There are many "generations" of computers to go before that point is reached! | |
|
|
To reach the point where machine translation equals in quality human translation, we would have first to be able to understand the mechanisms of human language, especially in terms of context, then explain it to the machine...
When the second point seems to be feasible in a foreseeable future, (and Google is apparently working on it), I believe we still have much to achieve on the first point.
translating is not like playing chess where the rules are clear and can be mathematically s... See more To reach the point where machine translation equals in quality human translation, we would have first to be able to understand the mechanisms of human language, especially in terms of context, then explain it to the machine...
When the second point seems to be feasible in a foreseeable future, (and Google is apparently working on it), I believe we still have much to achieve on the first point.
translating is not like playing chess where the rules are clear and can be mathematically schematized. A translator has to make decisions based on information and processes that nobody fully understands, that's where I see machine limitations, we cannot teach them what we don't understand.
As for Microsoft giving-up on it, it is a business decision, period. ▲ Collapse | | | l Gaston l Local time: 06:45 Afrikaan menyang Ibrani Kuno + ...
Ludivic, I agree 100% with you. | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Microsoft gives up on machine translation Trados Business Manager Lite |
---|
Create customer quotes and invoices from within Trados Studio
Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.
More info » |
| Wordfast Pro |
---|
Translation Memory Software for Any Platform
Exclusive discount for ProZ.com users!
Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |