Poll: "Machine learning will crack the code of language translation." Autor de la hebra: ProZ.com Staff
|
|---|
This forum topic is for the discussion of the poll question ""Machine learning will crack the code of language translation."".
This poll was originally submitted by Assem AlKhallouf. View the poll results »
| | | |
I've been involved in the development of machine translation since 1977--over 40 years. In the old days, the systems used linguistic rules. Then they tried building "statistical systems" based on machine learning. That didn't work, so now they're doing "hybrid systems" that combine both. It "works" up to a point for simple straightforward text. But the expectation that man will ever "crack the code" of true translation is preposterous. I get to see how laughable that assumption is every day.
... See more I've been involved in the development of machine translation since 1977--over 40 years. In the old days, the systems used linguistic rules. Then they tried building "statistical systems" based on machine learning. That didn't work, so now they're doing "hybrid systems" that combine both. It "works" up to a point for simple straightforward text. But the expectation that man will ever "crack the code" of true translation is preposterous. I get to see how laughable that assumption is every day.
This question is a far cry from "Will machine translation ever be useful for certain applications?" To which my answer is "Yes, it already has been for many years."
[Edited at 2018-01-29 08:32 GMT] ▲ Collapse
| | | | neilmac
España
Local time: 08:31
español al inglés
+ ...
I think "language translation" is about more than just cracking a code/s. And I like to cling to the human belief that there are some things that machines can't "learn"…
| | | | Nuno Rosalino
Reino Unido
Local time: 07:31
Miembro 2012
inglés al portugués
+ ...
I think the answer is an inevitable yes. Not a question of if, but when. But honestly, I'm not all that worried; when we have a general AI capable of translating at the level of a human professional, translation will not be the only, nor even the first, industry to fold. And it will probably still exist as an art, practiced by scholars for exegetical purposes. Like Tolkien said:
"The effort to translate, or to improve a translation, is valuable, not so much for the version it pro... See more I think the answer is an inevitable yes. Not a question of if, but when. But honestly, I'm not all that worried; when we have a general AI capable of translating at the level of a human professional, translation will not be the only, nor even the first, industry to fold. And it will probably still exist as an art, practiced by scholars for exegetical purposes. Like Tolkien said:
"The effort to translate, or to improve a translation, is valuable, not so much for the version it produces, as for the understanding of the original which it awakes."
[Edited at 2018-01-29 09:17 GMT] ▲ Collapse
| | |
|
|
|
| | | Maybe, however two trends must converge first | Jan 29, 2018 |
One trend, of course, is the development of machine translation to the point of extracting and processing the ideas behind the text, and no longer merely the words used to encode these ideas in a language, like it has been done to date.
It will be easy to notice when this time is approaching. Instead of only checking text for spelling and grammar, a text processing software - like Micro... See more One trend, of course, is the development of machine translation to the point of extracting and processing the ideas behind the text, and no longer merely the words used to encode these ideas in a language, like it has been done to date.
It will be easy to notice when this time is approaching. Instead of only checking text for spelling and grammar, a text processing software - like Microsoft Word - will be capable of checking it for consistency and logic too.
For instance, one could run a mystery book through it, and get feedback like, "Mr. Uptonogood could not be the hit-and-run driver on page 203, as he was arrested by police on page 187, and only released from custody on page 220.
Another example, running a sales report through it could result in something like "This cash flow figure (#1635) is inconsistent with the sales volume (#1292) and production cost (#792), after tax (#2087) has been paid, unless tax (#2087) was deferred, which should have been stated (#1778)."
Until this is no longer science fiction, machine translation will be limited to translating words and groups thereof. Evidence of this is that - at least for EN-PT - Google Translate has taken a major step back when it went "neural", as shown on http://www.lamensdorf.com.br/comparison.html
If machine translation could get any close to effectiveness, by now we'd have a 99.9% reliable automatic converter between Brazilian and European Portuguese. I have described some of the (tricky) issues at http://www.lamensdorf.com.br/ptxbr.html, showing that though it is (legally) one and the same language, it's not all about words and groups thereof.
The other trend is the gradual simplification of all languages, thanks to a combination of laziness and rush, leveraged by the growing use of somewhat uncomfortable onscreen keyboards on smartphones. As everyday language gets progressively depleted, it will require less computer resources (hardware & software) to "crack" human communication into bits and bytes.
A good illustration of this process was elicited in a joke recently posted by a friend of mine on Facebook:
"Where are you going, so circumspect, aloof, and unheeding?"
"I was going to take a crap, but now I'll go find a dictionary first." ▲ Collapse
| | | | | I don't think there is a code... | Jan 29, 2018 |
Basically, I think language is too fluid and irrational.
I strongly doubt the suggestion that languages are getting depleted.
The other trend is the gradual simplification of all languages, thanks to a combination of laziness and rush, leveraged by the growing use of somewhat uncomfortable onscreen keyboards on smartphones. As everyday language gets progressively depleted, it will require less computer resources (hardware & software) to "crack" human communication into bits and bytes.
A good illustration of this process was elicited in a joke recently posted by a friend of mine on Facebook:
"Where are you going, so circumspect, aloof, and unheeding?"
"I was going to take a crap, but now I'll go find a dictionary first."
First, people just jabbered.
Then, with great difficulty, they began recording their ideas with hieroglyphs carved in stone, or on clay tablets, or parchment...
Calligraphy developed, and more or less elaborate systems of writing, but all hard work.
And they preserved the most beautiful gems of language, all along.
Typewriters were a pain, but great authors managed to use them.
Texting with only a numeral keyboard must have been a real pain in the wrist and the thumb - I never mastered the art. But new, creative forms of language appeared, according to some researchers at least. The Danish language committee is delighted, because they do not regard texting as a degenerate form of Danish, but an elaboration of it. Maybe it was not all as artistic as haiku, but the adaptation to small screens and limited lengths of message forced users to develop codes that the receiver would understand... and the results were, from a linguistic point of view, highly creative. I am sure the same goes for all other languages.
Language develops, and is far more subtle than many people appreciate.
OK, we have trouble understanding anyone a couple of generations younger than ourselves, and it may be hard work reading literature from a couple of centuries back, but our language is not more primitive than the old classics. There are certainly some of the younger generation who are as creative as the best.
The last great authors have not been born yet.
I do not know where computers are going. But I do not believe all the science fiction. I think that for many generations to come, human logic will be too subtle for them. Humans will simply invent new codes, as they always have, and leave earlier generations (and computers) panting way behind.
| | | | |
|
|
|
All learning depends on the student's degree of motivation and commitment, particularly language learning.
A mechanised language-teaching system might well help a keen student - all forms of practice can be useful - but won't make any difference to an uninterested or de-motivated student.
| | | | Helen Hagon 
Local time: 07:31
Miembro 2011
ruso al inglés
+ ...
Machine translation will probably develop enough to take over the lower end of the market, although there will always be a need for people to make sure the machines are doing their job properly. However, I don't believe machines will ever develop empathy, understanding, or care about the quality of the end result, so where creativity, sensitivity and style matter, humans will still be required.
| | | | | The question wasn't about the market | Jan 29, 2018 |
I think some of you are missing the point. The question was whether **machine learning** can "crack the code" of the translation process. It's a very different question than what some of you have answered.
| | | | | Yes: one day | Jan 29, 2018 |
but it will take a while, even the neural translations are not even close yet...
| | |
|
|
|
John Fossey
Canadá
Local time: 02:31
Miembro 2008
francés al inglés
+ ...
| "crack the code"? | Jan 29, 2018 |
What does "crack the code" mean? And how would it be accurately translated into other languages? Perhaps some AI-powered MT engine could tell us?
| | | | DZiW (X)
Ucrania
inglés al ruso
+ ...
| Verisimilitude | Jan 29, 2018 |
Retrieve, accumulate, classify, select, backtrack--and properly use the 'human' weighted patterns and exclusions... just a piece of stereotype!
According to rather many papers since 1969 like this, the only corner feature (but irrationality) the machine lacks is hominization via the <... See more Retrieve, accumulate, classify, select, backtrack--and properly use the 'human' weighted patterns and exclusions... just a piece of stereotype!
According to rather many papers since 1969 like this, the only corner feature (but irrationality) the machine lacks is hominization via the bias. Makes sense and requires not much to bruteforce nowadays!
That's it: the common feature of the variety of people (aka mankind) is but prejudice ▲ Collapse
| | | | | Not at all. The Watson experience is the best example! | Jan 29, 2018 |
José Henrique Lamensdorf wrote:
(...) Until this is no longer science fiction, machine translation will be limited to translating words and groups thereof. Evidence of this is that - at least for EN-PT - Google Translate has taken a major step back when it went "neural", (...) If machine translation could get any close to effectiveness, by now we'd have a 99.9% reliable automatic converter between Brazilian and European Portuguese. (...) The other trend is the gradual simplification of all languages, thanks to a combination of laziness and rush, leveraged by the growing use of somewhat uncomfortable onscreen keyboards on smartphones. As everyday language gets progressively depleted, it will require less computer resources (hardware & software) to "crack" human communication into bits and bytes.
Google translate, Facebook & Twitter with Bing Translate and further more "Neural" apps created to be a kind of AI for Portuguese (at least my Portuguese from Portugal) will not prevail unless the Human Translators let that happen.
The question is: are we prepared to shut down the fake "neural"?
«Most technology reinforces the feeling that reality is just a sea of gadgets; your brain and your phone and the cloud computing service all merging into one superbrain. You talk to Siri or Cortana as if they were people. VR is the technology that instead highlights the existence of your subjective experience. It proves you are real.» Jaron Lanier at http://www.jaronlanier.com/dawn/
| | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Poll: "Machine learning will crack the code of language translation." | Pastey | Your smart companion app
Pastey is an innovative desktop application that bridges the gap between human expertise and artificial intelligence. With intuitive keyboard shortcuts, Pastey transforms your source text into AI-powered draft translations.
Find out more » |
| | 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... | | | | | |
Login to reply/comment 
