Pages in topic:   [1 2] >
Machine translation, again
Thread poster: Claudio Porcellana
Dec 10, 2009

do a look to this intersting link
http://portal.acm.org/citation.cfm?id=1562764.1562798

it seems that MT is not born to reduce costs/rates, but rather as expert translators are lacking ...
it makes one think, isn't it?

and here is the link to the whole article
http://rali.iro.umontreal.ca/Transtype2/TT2-CACM-Oct09.pdf

very intersting this Wikipedia page
http://en.wikipedia.org/wiki/Machine_translation
and above all the part about Claude Piron

Claudio

[Modificato alle 2009-12-10 17:50 GMT]


Direct link Reply with quote
 

Richard Bartholomew  Identity Verified
Germany
Local time: 01:09
Member (2007)
German to English
This approach has some merit. Dec 13, 2009

Yes, Mr. Porcellana, the article you refer to does make me think - and I find that to be a good thing.

While I don't think the approach described will negatively impact the translation business from the human translator's viewpoint, it might positively influence MT research. The little reading I've done on the subject indicates that most current MT research revolves around corpora of already translated pairs.

Collecting keystroke-by-keystroke data about how human translators generate the translated strings observed in a corpus might help researchers to shed light on how human translators translate, not just what the end product is. It's the difference between seeing the end result and seeing how it happens.

As I understand it, contemporary MT research is divided between the rule-based and statistical camps. Do human translators use rule-based or statical strategies when they translate? Or do they use some combination? Or do they have some other, completely different, strategy not yet investigated? I wouldn't mind finding out. How about you?


Direct link Reply with quote
 

Sheila Wilson  Identity Verified
Spain
Local time: 00:09
Member (2007)
English
+ ...
I don't think we use A (i.e. one) strategy Dec 13, 2009

Richard Bartholomew wrote:
As I understand it, contemporary MT research is divided between the rule-based and statistical camps. Do human translators use rule-based or statical strategies when they translate? Or do they use some combination? Or do they have some other, completely different, strategy not yet investigated? I wouldn't mind finding out. How about you?


I somehow doubt that they are going to find that translations (at least, French to English ones) follow any particular rule or even combination of rules. Isn't that what sets us apart from MT?

As you say, I wouldn't mind finding out - as long as they don't find out how to replicate it!


Direct link Reply with quote
 

Richard Bartholomew  Identity Verified
Germany
Local time: 01:09
Member (2007)
German to English
Beat 'em or join 'em Dec 14, 2009

Well, Ms. Wilson, if they do find a way to get machines to credibly replicate human performance, which I doubt they will any time soon, if ever, I guess we could try to join the MT crowd. But for me, that would mean getting involved with the very two technical subjects I'm least fond of: logic programming a la Prolog on the rule-based side and statistics on the SMT side. That's the kind of Scylla and Charybdis situation I'd prefer to avoid if possible.

Some introspection is useful as far as understanding how human translators translate. Roughly speaking, I'd say that 5% of what I do is rule-based and 5% statistical. A lot of the job is just chopping up source language sentences, translating the pieces, and rearranging the translated pieces into a target sentence. That leaves a lot of noman's land between what I do and what current MT programs do. I'll be content if it stays that way for another ten or fifteen years.


Direct link Reply with quote
 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 01:09
Member (2005)
English to Spanish
+ ...
Translators caught in off-side Dec 14, 2009

I think that the problem is that we might be caught in an off-side position in the long run. Let me explain: does it not happen in your country that politicians say that they will reduce unemployment, curtail unproductive public expense, etc. etc., and when you examine the figures after some time these things haven't happened at all, but instead they have altered the way these figures are calculated and published, so that --on the paper at least-- they did honour their commitment?

The same could be happening with translation. Those interested in saving money with automatic translation may be trying to alter the consciousness among the public about what is a text with a good quality, with the goal of making automatic translation pass as good translation. This does not bother me much, but it does bother me that, in the future, we translators use googled automatic translations as a reference for our work. Terrifying, huh?


Direct link Reply with quote
 
Laurent KRAULAND  Identity Verified
France
Local time: 01:09
French to German
+ ...
OT: conditions already met Dec 14, 2009

Tomás Cano Binder, CT wrote:
The same could be happening with translation. Those interested in saving money with automatic translation may be trying to alter the consciousness among the public about what is a text with a good quality, with the goal of making automatic translation pass as good translation. This does not bother me much, but it does bother me that, in the future, we translators use googled automatic translations as a reference for our work. Terrifying, huh?

Hi Tomás,
from what I see on Francophone fora here and there, the prerequisites for such a paradigmal shift are already met. People tend to write and to express themselves according to random rules or by trusting their "feeling" about how words and sentences should be written.
There was also a recurring discussion in the media some months ago, some thinking heads saying that reforming French through simplification (as it already is done in Canada) would help bridge the societal gap. As a consequence, standard French should be reserved for formal use.
What those thinking heads failed to see is that (at least according to my "feeling") such a change would only widen the societal gap.
In my opinion, the same would happen with translation - that is, if automatic translation devices are able to decipher a phonetically or randomly written language (and if people writing phonetically or randomly still have the intellect to understand formal language).
The idea that a majority is right simply because it is the majority is a deadly way of thinking. And we should never forget that, for better and for worse, democracy way too often goes hand in hand with demagogy.

[Edited at 2009-12-14 08:04 GMT]


Direct link Reply with quote
 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 01:09
Member (2005)
English to Spanish
+ ...
Slackening education widens the social gap Dec 14, 2009

Laurent KRAULAND wrote:
There was also a recurring discussion in the media some months ago, some thinking heads saying that reforming French through simplification (as it already is done in Canada) would help bridge the societal gap. As a consequence, standard French should be reserved for formal use.
What those thinking heads failed to see is that (at least according to my "feeling") such a change would only widen the societal gap.

I completely and entirely agree. Proof that lowering the cultural and educational requirements is a big mistake is what is happening in Spain. Some 15 years ago the education system was softened to make education compulsary up to 16 years of age. This meant that kids quite simply could not be thrown out of school until they were 16 years old, no matter how badly they fared as students or how badly the behaved in class. The result? Now hardly 10% of kids are minimally prepared for higher education, and in classes abut 50% of pupils simply don't care a damn about their education and consistently get bad marks. I see no future for them apart from working as shop attendants and cheap labor in warehouses. Their parents feel that if their children get bad marks is because teachers are unable to motivate the class.

After they finish intimidating teachers, parents take their kids to the next department store and get the kid the latest videogame system, their favourite clothes, their own TV set, or a computer so that they can socialise from their room. Being the husband of a teacher, I hear this from my wife and her colleagues every day.

The previous system, which was in place for many decades with just minor alterations, had really good results: a majority of entrepreneurs and people in managerial levels today in Spain come from very normal, blue-collar, and farming families. I am pretty sure we will not be able to say that in 10 years time, as only people with a pile of money can afford a better, stricter education for their children in private schools.

We definitely need a change in Spain, but no politician seems to want to take the risk of not being voted by young people, who would see a stricter education as some kind of dictatorial attitude.

[Edited at 2009-12-14 08:40 GMT]


Direct link Reply with quote
 

david young  Identity Verified
France
Local time: 01:09
French to English
it will happen Dec 14, 2009

"Well, Ms. Wilson, if they do find a way to get machines to credibly replicate human performance, which I doubt they will any time soon, if ever"

If you accept that the human brain is a machine, albeit a warm and wet one, then machine will catch up one

day, and once they've caught up they'll shoot ahead of us in the space of a few years.

You might find this site interesting.
http://www.kurzweilai.net/index.html?flash=1


Our only solace is that language will the last nut that machine intelligence will crack...

[Edited at 2009-12-14 09:27 GMT]


Direct link Reply with quote
 
Laurent KRAULAND  Identity Verified
France
Local time: 01:09
French to German
+ ...
Fascination for machines Dec 14, 2009

david young wrote:

You might find this site interesting.
http://www.kurzweilai.net/index.html?flash=1

Our only solace is that language will the last nut that machine intelligence will crack...

[Edited at 2009-12-14 09:27 GMT]


I never quite understood mankind's fascination for machines which (will) have the potential to make mankind redundant in every respect.


Direct link Reply with quote
 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 01:09
Member (2005)
English to Spanish
+ ...
Are we emulating Gordias? Dec 14, 2009

david young wrote:
Our only solace is that language will the last nut that machine intelligence will crack...

It looks to me that we could be as naïve as Gordias and that sheer computer power and database technology could be the Alexander the Great that cuts the knot.

We should not expect a computer to be intelligent --AI is sort of démodé anyway nowadays--, but we can surely expect statistical features to solve most of easy translations in the medium run.

I have said this many times: those who want to be translators in 10 years time must prepare, get certified in one or more schemes, learn, and strive for perfection in each and every job. Unless we continue to become better day after day and prove that we are systematically superior, computers will swallow a big share of our work at some stage.


Direct link Reply with quote
 
Laurent KRAULAND  Identity Verified
France
Local time: 01:09
French to German
+ ...
Subtle nuance Dec 14, 2009

Tomás Cano Binder, CT wrote:
It looks to me that we could be as naïve as Gordias and that sheer computer power and database technology could be the Alexander the Great that cuts the knot.

Although I agree with you on that point, Tomás, I would like to make a subtle nuance here: cutting the Gordian knot in two is not untying it (at least not technically/essentially seen). And I would really be suspicious towards those who think/say: "Who cares? Only the result matters"... this for reasons I feel do not need to be detailed.


Direct link Reply with quote
 

KSL Berlin  Identity Verified
Portugal
Local time: 00:09
Member (2003)
German to English
+ ...
Careful with your assumptions Dec 14, 2009

david young wrote:
If you accept that the human brain is a machine, albeit a warm and wet one, then machine will catch up one day, and once they've caught up they'll shoot ahead of us in the space of a few years.


Ah, but I don't. Biological entities are so much more subtle and complex, and the notion that one can truly reproduce complex thought processes through programmatic modeling is naive and/or presumptuous. That mimicking at a primitive level is clear from the "success" of various current MT engines, but there is no simple linear path from there to the sophisticated use of language. And as much as "controlled language" is talked about, I have seen few examples that indicate there is any hope of its consistent application in technical writing.

Laurent KRAULAND wrote:
What those thinking heads failed to see is that (at least according to my "feeling") such a change would only widen the societal gap.


Quite true. The same applied to the "ebonics" movement in the US and US attempts at bilingual education, which are largely a swindle that ruins the career chances of the children later in life. Lest anyone think I'm slamming bilingualism here, I'm not. Bilingual education (Spanish/English) in the US simply ignored some basic issues like the fact that many of those segregated in its programs did not come from literate backgrounds in their native language and those who did had no need of such programs to succeed academically. Trying to achieve literacy first in the language which is not dominant in the wider culture proved in practice to be a very bad approach no matter what some theories have to say about the matter.

In any case, as I've said before, the only translators who need to fear MT are those who should probably be digging ditches and washing dishes now anyway. The technology will, at best, be only an unreliable assistive tool in our lifetimes.


Direct link Reply with quote
 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 01:09
Member (2005)
English to Spanish
+ ...
We do care, but that is not enough Dec 14, 2009

Laurent KRAULAND wrote:
Tomás Cano Binder, CT wrote:
It looks to me that we could be as naïve as Gordias and that sheer computer power and database technology could be the Alexander the Great that cuts the knot.

Although I agree with you on that point, Tomás, I would like to make a subtle nuance here: cutting the Gordian knot in two is not untying it (at least not technically/essentially seen). And I would really be suspicious towards those who think/say: "Who cares? Only the result matters"... this for reasons I feel do not need to be detailed.

Well, we translators would certainly like to see computers untying the knot, but in this case what corporations want is some translation and they will not care if the knot was cut or untied.

Our only power against machine translation is good translations time and time again. Let's strive for that with training, certification, and good work.


Direct link Reply with quote
 
Claudio Porcellana  Identity Verified
Italy
TOPIC STARTER
Machine translation, again Dec 14, 2009

thanks one and all for this brainstorming


indeed I had no intention to debate about MT reliability (that is currently 2, to me, on a scale from 1 to 10), but rather to think about the suggestion hidden in this article, i.e. the fact that specialization and improvement are the only way to beat (in the middle or in the long period) MTs, as somebody stated (Tomas, among the others)

as a matter of fact, these scientists didn't use, as reasons for developping TMs, the more common and defensible causes, i.e. cost & speed, but rather the lack of skilled translators! and I think too it makes sense

As an example, few days ago I received the request to translate a very difficul military document
As I was able to understand, my peer had many difficulties to find a team
I refused the proposal as the deadline was too tight for such a complex topic, but I immediately thought to this thread: indeed I understand that it is not so easy to find a team of translators that know well enough certain topics, so I can understand the temptation to use MT when you can't find it

Claudio

Tomas said:
I have said this many times: those who want to be
translators in 10 years time must prepare, get certified in
one or more schemes, learn, and strive for perfection in
each and every job. Unless we continue to become better day
after day and prove that we are systematically superior,
computers will swallow a big share of our work at some
stage.


Direct link Reply with quote
 
Laurent KRAULAND  Identity Verified
France
Local time: 01:09
French to German
+ ...
A guideline: Dec 14, 2009

Make it "perfect" the first time... every time (although I fail to remember who coined this sentence).

Tomás Cano Binder, CT wrote:

Well, we translators would certainly like to see computers untying the knot, but in this case what corporations want is some translation and they will not care if the knot was cut or untied.

Our only power against machine translation is good translations time and time again. Let's strive for that with training, certification, and good work.

Exactly.


[Edited at 2009-12-14 10:16 GMT]


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 »

Machine translation, again

Advanced search







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 »
TM-Town
Manage your TMs and Terms ... and boost your translation business

Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.

More info »



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