Mobile menu

Pages in topic:   [1 2] >
Off topic: Google - the future of translation?
Thread poster: Carolingua

Carolingua  Identity Verified
United States
Local time: 21:57
Spanish to English
+ ...
Feb 8, 2005

I recently saw a TV interview with one of Google's founders, who said that the company is working on developing/perfecting machine translation...I think they may even have said this was a strategic priority for them, which stuck in my mind and got me thinking. Machine translations are far from perfect at the moment, because the software seems unable to accurately discern context. On the other hand, Google is a highly-successful company with extremely bright employees--it seems to me if they set their sights on something they may actually have a shot at achieving it. Do you think that given sufficient power and data, a computer could learn to interpret the context and eventually take the place of a translator? If so, are any of you worried about the future of the translation industry?

I'm interested to know if any of you had heard about this (i.e. Google's intentions), and to hear your opinions about whether or not it's true and discuss the issue and the questions it raises (i.e. is it possible to develop software that translates as well as a human? is it desirable? what should we be doing as translators to ensure we remain marketable in the future?)


[Subject edited by staff or moderator 2005-02-09 00:28]


Direct link Reply with quote
 

Uldis Liepkalns  Identity Verified
Latvia
Local time: 07:57
Member (2003)
English to Latvian
+ ...
I don't believe in foreseeable future Feb 8, 2005

machines will substitute humans. The main reason of such my belief is that we do not translate words or sentences, we translate IDEAS.

Uldis


Direct link Reply with quote
 

Juan Pablo Solvez Beneyto  Identity Verified
Spain
Local time: 06:57
Member (2002)
English to Spanish
+ ...
Google's opinion counts Feb 8, 2005

Here it is:

http://www.google.com/intl/en_ALL/help/faq_translation.html

"The translation isn't as good as I'd like it to be. Can you make it more accurate?

The translation you are seeing was produced automatically by state-of-the-art technology. Unfortunately, today's most sophisticated software doesn't approach the fluency of a native speaker or possess the skill of a professional translator. Automatic translation is very difficult, as the meaning of words depends upon the context in which they are used. Because of this, accurate translation requires an understanding of context, as well as an understanding of the structure and rules of a language. While many engineers and linguists are working on the problem, it will be some time before anyone can offer a quick and seamless translation experience. In the interim, we hope the service we provide is useful for most purposes."

As for my opinion, I mostly agree with the above.

Regards


Direct link Reply with quote
 

Selçuk Budak  Identity Verified
Local time: 07:57
English to Turkish
+ ...
Invisible Lunatic Feb 9, 2005

The problem of intelligent machines is as hot as ever since the time of Alan Turing, a talented English mathematician with a really unfortunate fate. He proposed a machine now called Turing machine and a test procedure called Turing Test. At the beginning, the theory was welcome by many scientists studying cognition and artificial intelligence. But then arose serious criticisms, as for example put forth with Chinese Room (Searle). Those colleagues interested in arguments as to why computer “intelligence” cannot not replace human cognition can google “Chinese room” and read about it.

I remember from my Cognitive Psych courses an anecdote that seems pertinent here: Scientist had devised an intelligent machine capable of making translations and wanted to test it with English –> Chinese –> English. If, they assumed, they could get the original sentence after back translation, then this means that the machine is intelligent and successful. So they fed it an English idiom: “Out of mind out of sight.” The machine first translated it into Chinese, and then back into English. The result was: “Invisible lunatic.” A perfect dictionary translation, indeed. “Out of mind = lunatic” and “out of sight” = invisible.

An ironic, similar anecdote is told in Turkey as a joke. Scientists invented a computer which is as much as, or more intelligent than, a human being. And they challenged, and claimed that the computer is capable of answering all kinds of questions conceivable by human mind. An ignorant guy from Turkey came and asked: “Ne iºti?” A silly question indeed. It means, depending on the asker’s intention and background, as well as on the context: “How are you? Or “What did s/he drink?” Or “What the hell you doing?” The computer tried hard to find an answer, failed, and blew out. Implied lesson to be learnt, so tellers of this joke say, that even the most ignorant guy is smarter than the most sophisticated computer ever conceivable by human imagination. Because, the computer is not, and cannot be, intelligence, not cognition, not intuition, but only a shadow, an imitation, or a diminution of it. The intuition which the computer lacks, is a crucial factor in contextual perception of language. The computer, as proponents of Chinese Room criticism assert, is capable of processing syntax, but it is not capable of interpreting semantics.

In essence, the question raised by Carolingua is identical with the question raised about the CAT tools that already caused considerable uneasiness among our colleagues. And it has some objective truth in reality. A simple example:
Last year, I translated an operating manual of around 60,000 words. Early this year, I translated its new edition with extensive revision: Only 25,000 words. Next year, it would perhaps be diminished 10-15 thousand words. In one sense, this means a real loss. Even, a Turkish colleague insists, on various forums, that the translators should not surrender their TM’s to clients.

The uneasiness that we feel about CATs, googles, etc. is not a concern for the translator of Ulysses or Homer. We translate operating manuals, instructions, agreements, etc. with highly structured formats, simple language, and a narrow context with definite terminology. Therefore, when objectively considered, it is certain that machine translation would replace much human translation with awesome TM and processing capabilities and speed.

This is a phenomenon, and we have already begun to experience its consequences as illustrated by the example mentioned above.

But, does this mean that in the foreseeable future we, translators, would be deprived of our jobs? Definitely not. Today, CAT tools are around everywhere, and we survive our jobs. Even the better, those of us with CAT experience seem to be more advantageous with regard to receiving translation assignments.

If not, then what does this mean? It only means that we should adapt our skills and our orientation to new developments and new circumstances. Those of us above 40s would get what I mean. We began doing translation with pen. Then we switched to mechanical typewriters, then to first PCs with only 64K RAM and 180K single-sided floppies without hard disks at all. Now we work with Gigabytes and CATs and googles, and vast online resources. And thanks to internet, we know, that our work is much easier, much safer, and much accurate and reliable today than ever.

We should treat this phenomenon not as our rival, but as a very useful tool that we can exploit to our advantage. As a further reassurance, we should bear in mind that today’s CAT users are highly capable translators, and not monolingual employees. This, in itself, is the most striking evidence that even the most sophisticated CAT tool with huge capabilities would need bilingual persons like us to operate!

Therefore, please keep smiling.

SB.


Direct link Reply with quote
 
pep
Local time: 06:57
English to Spanish
There is a risk, comparable to translating with non-native translators Feb 9, 2005

My opinion is that with machine translation there is an added risk, comparable to translating with non-native translators, and cannot be used with certain types of translations.

If you use the google translation: http://translate.google.com/translate_t

to translate "My Mom is cool and nice" to Spanish and then back to English, you get at return: "My breast is fresh and pleasant".

Pep.


Direct link Reply with quote
 

Jabberwock  Identity Verified
Poland
Local time: 06:57
Member (2004)
English to Polish
I don't share your optimism... Feb 9, 2005

The arguments which support the claim that the machine cannot ever translate texts like a human are very similar to those given for the chess computers. People said that a computer cannot think, cannot improvise tactics, cannot intuitively select the best solution etc.

Unfortunately, it turned out that it does not have to. The mere computing power is enough to apply a "brute force" approach. Sure, it lacks finesse, is not that elegant, but it allows the computer to win with most of the average players and some of the champions.

That may be the case with computer translation as well. Sure, the computer will not translate poetry well, but it may turn out that, given enough input data, it will produce acceptable technical manuals. Not the best, not elegant, with some minor mistakes or awkward phrases, but acceptable and ten times cheaper than a translation made by a human.

It will not happen any time soon, but in ten years I think it would be wise to take it into account.


Direct link Reply with quote
 

Stefano Papaleo  Identity Verified
Italy
Local time: 06:57
Member (2005)
English to Italian
+ ...
man vs. machine Feb 9, 2005

I agree that there's not going to be a good MT software for a long, long time, simply because computers are fast but are not smart at all and maybe they will never be. We know very little or nothing at all about our brain and how we learn languages (esp. L1) so it's going to take a while before we can "teach" this to a machine.
Context is the key factor, and that's why most of the automatic translations are just nice funny jokes.
Yet, someone raised a good point: chess. Sure, computers are good at that and other games, they were created to process data faster and better than any human could ever dream of, and that's what they do. So, probably given a little time, resources, automatic translations will take over in some fields, simply because as it has been pointed out, a manual or an agreement or even a web page is not a novel or a poem. Bear in mind also that they will be cheap, and clients are veeeery concerned about money, often more than they are about quality, elegance etc.
It happened before in almost any sector of human activities, computer can paint or write music.. of course it won't be a Raffaello or a Beethoven but... art is art and business is business.
Maybe we will be hired no longer as translator but as advisors or project managers for these applications, I don't know. And, yes... clients should pay a good extra for our TM's, just like we pay royalties when we buy a cd;)


Direct link Reply with quote
 

lien
Netherlands
Local time: 06:57
English to French
+ ...
No fear Feb 9, 2005

As long a machine will translate :

"You have this groovy little sports car ripe for an eighteen- to thirty-year-old demographic."

in

"Vous avez cette petite voiture de sport routinière mûre pour des dix-huit à trente-année-vieux démographique"

there are still sunny days for translators.

The machine will never remplace the translator, because translating is to convey a message, understanding what the author wants to say beyond the plain meaning of the words, which requires a feeling backed by the own life experience of the person behind the translator.


Direct link Reply with quote
 

Wouter van Kampen
Thailand
Local time: 11:57
Danish to Dutch
+ ...
Bad news for most of you Feb 9, 2005

The Hilbert Engine from Primentia Inc. (www.primentia.com)
turning databases into multi-dimensional vector arrays. Vector algebra is going to be the winning game. Computers have great difficulty processing text since going through all possible combinations of text strings previously stored in a TM and working out the most likely proper combination given the context of the original string within an acceptable time frame is quite an impossible task for todays CAT-tools. Things get really different when CAT-engine builders decide to leave the old database technology behind and turn to the vector approach.

So be prepared. Translation of legal documents by humans will soon be history. Technical translations await the same fate.

Think twice when you consider starting of a career as a technical translator. The only work that will be left for you is translating the kind of documents that are so poorly written that only the ones that really understand what the original writer had in mind to convey, will be able to rewrite the contents in another language. And that my dear friends will put the majority of technical translators out of business.

Wouter van Kampen
(Feb. 9, 2005)


Jabberwock wrote:
--snip--
Sure, the computer will not translate poetry well, but it may turn out that, given enough input data, it will produce acceptable technical manuals.
--snip--
It will not happen any time soon, but in ten years I think it would be wise to take it into account.


Direct link Reply with quote
 

Jens Mährländer  Identity Verified
Local time: 06:57
English to German
+ ...
It is already happening Feb 9, 2005

During the last 7 or 8 years I made 80 % of my annual turnover translating articles for the Microsoft Knowledge Base (English - German). After a test phase in Spain, where customers where reported to be more or less satisfied with the machine translations, Microsoft adopted the same system in Germany as of July 2004 cutting down the volume left for human translators by 75%. The quality of these machine translations is akward, but should the German customers behave in the same way as there Spanish counterparts, MS is going to stick to this theme to cut down on costs. If you happen to see such a machine translated article, you are encouraged to report how satisfied you are with the translation (on a scale from 1 to 9 I think). Well, you will know what you have to do to save our profession (at least for ten more years or so).

Direct link Reply with quote
 
Aleksandr Okunev
Local time: 07:57
English to Russian
Driverless cars are in the news every month Feb 10, 2005

Jens Mährländer wrote: to save our profession...

I use one of the best MT items on a regular basis, have it hooked to Wordfast. It can churn out a 100% correct segment, like "Open a new file and edit it", once in a while, but mostly I use it as a huge customizable dictionary, a sort of terminology server. I do not think that MT and artificial intelligence are different things, as some folks claim, to work in the real world you simply cannot avoid being human.


Direct link Reply with quote
 

Patrice  Identity Verified
United States
Local time: 21:57
Member
French to English
+ ...
They wish! Feb 14, 2005

I use machine translation for a good laugh. The nonsense produced is hysterical. I suppose theoretically machine translation could be improved upon, but it would take so much research and and development and I wonder if Google is that committed to the idea. I agree with the person who responded that we translate ideas rather than words. If Google wants to pour money into R&D to deal with semantics etc at a very high level...well, not to worry. They will have to hire many people like us to work on those projects.

My biggest concern about the translation profession is the constant pressure to work for less money. I am surprised at how little some of my colleagues are apparently willing to charge for their knowledge and hard work. TM has had a negative effect on rates. Trados sells to translators with the idea that they need never translate the same sentence twice and the Trados website indicates that they are creating demand for translators by convincing companies of the need for global translation. True. However, they must also be somehow making clear to corporate buyers of Trados that their need for translators will be reduced. I already observe that companies using TM want to pay less for word matches etc. I don't mean to single out Trados. This phenomenon applies to TM in general. I think it is important for translators to hold the line on how low they are willing to go, rate-wise.


Direct link Reply with quote
 

Jeff Allen  Identity Verified
France
Local time: 06:57
Member (2011)
Multiplelanguages
+ ...
Google's Machine Translation engine(s) in Google Tools Mar 5, 2005

Carolingua wrote:
I recently saw a TV interview with one of Google's founders, who said that the company is working on developing/perfecting machine translation...I think they may even have said this was a strategic priority for them, which stuck in my mind and got me thinking. Machine translations are far from perfect at the moment, because the software seems unable to accurately discern context. On the other hand, Google is a highly-successful company with extremely bright employees--it seems to me if they set their sights on something they may actually have a shot at achieving it. Do you think that given sufficient power and data, a computer could learn to interpret the context and eventually take the place of a translator? If so, are any of you worried about the future of the translation industry?

I'm interested to know if any of you had heard about this (i.e. Google's intentions), and to hear your opinions about whether or not it's true and discuss the issue and the questions it raises (i.e. is it possible to develop software that translates as well as a human? is it desirable? what should we be doing as translators to ensure we remain marketable in the future?)


Google is not itself doing any work on machine translation development. Just like all of the other search engine portals, they are taking existing MT engines and making them available through the Google Tools portal. With a little background research, it is possible in 30 minutes to discover what engine(s) they are using.

I've described this type of approach in:
ALLEN, Jeff. 2000. The Value of Internet Translation Portals. In International Journal for Language and Documentation (IJLD), Issue 6, August/September 2000, pp. 45-46.
http://www.geocities.com/mtpostediting/ijld6.doc

As for discerning context, it depends which MT system or software package you are using. I test and use most of the commercially available MT software. There are some that are able to deal with contextual analysis.

Examples of the results of my real translation work resulting from using MT are available:

ALLEN, Jeff. 2001. Postediting: an integrated part of a translation software program (Reverso Pro). In Language International magazine, April 2001, Vol. 13, No. 2, pp. 26-29.
http://www.geocities.com/mtpostediting/Allen-LI-article-Reverso.pdf

ALLEN, Jeff. MT User case study in the Telecom field: pre-sales and post-sales documentation. Presented at AMTA2004, Sept-Oct 2004.
http://www.geocities.com/mtpostediting/Jeff-Allen-AMTA2004-paper_v1.01.pdf

& a new set of results in:

ALLEN, Jeff. What is Post-editing? Translation Automation Newsletter, Issue 4. February 2005.
http://www.geocities.com/mtpostediting/TA_IssueFour.pdf


Several reviews of several commercial software programs are:

Reverso 5 (PROMT v5)
http://www.multilingual.com/allen50.htm
SYSTRAN v4
http://www.multilingual.com/allen58.htm
PROMT XT (v6) 2003
http://www.multilingual.com/allenWassmer62.htm
Pocket PROMT 4.0
http://www.multilingual.com/allen68.htm


Jeff
http://www.geocities.com/mtpostediting/


[Edited at 2005-03-17 12:46]

[Edited at 2005-06-14 18:09]


Direct link Reply with quote
 

Jeff Allen  Identity Verified
France
Local time: 06:57
Member (2011)
Multiplelanguages
+ ...
Microsoft Data-driven MT project Mar 5, 2005

Jens Mährländer wrote:
During the last 7 or 8 years I made 80 % of my annual turnover translating articles for the Microsoft Knowledge Base (English - German). After a test phase in Spain, where customers where reported to be more or less satisfied with the machine translations, Microsoft adopted the same system in Germany as of July 2004 cutting down the volume left for human translators by 75%. The quality of these machine translations is akward, but should the German customers behave in the same way as there Spanish counterparts, MS is going to stick to this theme to cut down on costs. If you happen to see such a machine translated article, you are encouraged to report how satisfied you are with the translation (on a scale from 1 to 9 I think). Well, you will know what you have to do to save our profession (at least for ten more years or so).


This is part of Microsoft huge project for its Data-Drive MT systems (DDMT). It is basically a rebaptized name for Example-Based MT.

The Natural Language Processing (NLP) group at Microsoft has presented a couple of conference presentations on their DDMT work in the past couple of years.

Jeff
http://www.geocities.com/mtpostediting/


Direct link Reply with quote
 
Mohamed  Identity Verified
United States
Local time: 22:57
English to Arabic
+ ...
TAUS, IBM and LionBridge Partnership, Google, etc May 6, 2010

Yes, I am sure this will have an impact on our business. Machine translation is a hot topic these days and it seems everybody is talking about it. TAUS is leading the wave and their goal is to bring all TMs form big players (Microsoft, Oracle, SDL, etc) and make them available for free. There will be a subscription fee as well which will give you more.

IBM and LionBridge are now partners and IBM will get its machine translation into play with the help of LionBridge.

Our job will change dramatically in the future. Most probably, our translation will be replaced with 'post-editing'. This is ongoing as we speak/write. Read this please to see what I am talking about "http://blog.gts-translation.com/2010/03/03/post-editing-machine-translation-using-systran-and-trados/"


The most important question now is: what can we do as translators? We really need to answer this. Should Proz stand and do something about it? I am not sure. Big players are leading the wave for machine translation and I am not sure if we, poor translators, can do something or we should just ride the wave. This is a tough question that pops up in my mind right now.

What we can do:

1) Do not accept post-editing jobs
2) If you accept post-editing jobs, charge higher rates
3) Translators should set a threshold for a post-editing rate that is fair for translators and clients in the same time


Direct link Reply with quote
 
Pages in topic:   [1 2] >


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


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

Google - the future of translation?

Advanced search


Translation news





PerfectIt consistency checker
Faster Checking, Greater Accuracy

PerfectIt helps deliver error-free documents. It improves consistency, ensures quality and helps to enforce style guides. It’s a powerful tool for pro users, and comes with the assurance of a 30-day money back guarantee.

More info »
Protemos translation business management system
Create your account in minutes, and start working! 3-month trial for agencies, and free for freelancers!

The system lets you keep client/vendor database, with contacts and rates, manage projects and assign jobs to vendors, issue invoices, track payments, store and manage project files, generate business reports on turnover profit per client/manager etc.

More info »



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