Pages in topic:   [1 2 3] >
Do you see Systran as an aid or as competition?
Thread poster: Andrey Lipattsev

Andrey Lipattsev  Identity Verified
Ireland
Local time: 00:53
English to Russian
+ ...
Oct 16, 2006

After reading a most stimulating thread just now on the ethics and economics of using CAT, I thought it would be interesting to discuss the issue of machine translation.
I find using Systran a lot more helpful than using traditional CAT (Trados, DejaVu, etc.)
Traditional CAT just slows me down most of the time - some clients like Trados, some SDLx, some DejaVu, etc. I can't really spend time to learn how to use all these tools to their full extent. As a result, whenever I'm asked to use a specific CAT tool, I take the job (not rich enough yet to throw away potential orders ), but I end up spending more time on it (sometimes for a reduced/discounted rate) then I would doing it all manually.
With Systran, I save myself a lot of time on dictionary search, on ensuring consistency and so on. Even if the client has a preferred glossary of terms, find&replace does the job once I'm finished translating my own way.
What do you people think? Is Systran a replacement for us, humanoid translators, to be feared or is it a powerful tool that allows us to do our job quicker and better?


Direct link Reply with quote
 

texjax DDS PhD  Identity Verified
Local time: 18:53
Member (2006)
English to Italian
+ ...
look at this! Oct 16, 2006

When I have read yours thread, for a moment I have had the doubt that be a matter of one joke. I had not never felt to name Systran, I notice hour that is equivalent to Babelfish.
Therefore I have decided to write my participation in Italian (mine madrelingua) and of sottoporlo to “feared” Systra. Cio' that you read of it e' the result. Therefore, in order to answer to your question:

Andrey Lipattsev wrote:
What do you people think? Is Systran a replacement for us, humanoid translators, to be feared or is it a powerful tool that allows us to do our job quicker and better?


Not, not creed that is null to fear.

-End of Systran translation-

No further comments


Direct link Reply with quote
 

Andrey Lipattsev  Identity Verified
Ireland
Local time: 00:53
English to Russian
+ ...
TOPIC STARTER
a bit more research Oct 16, 2006

well, thanks for that rapid if somewhat rash debunking, texjax, but you may want to spend a bit more time on researching this issue furhter.
babelfish is, indeed, powered by Systran. i don't see how that makes them "equivalent" though. Altavista's babelfish is a fun, stripped down version of the full program. hence, the question still stands.
indeed, if you look at your babelfish translation, it is not perfect, but we can surely all understand it and I could make a normal text out of it in 5 minutes, whereas if you gave me your Italian original, it would take me considerably longer to translate it (I do not speak Italian very well, but I do have some notions of it)

[Редактировалось 2006-10-16 21:03]


Direct link Reply with quote
 

Rafa Lombardino
United States
Local time: 15:53
Member (2005)
English to Portuguese
+ ...
No, no, no... Oct 16, 2006

Andrey Lipattsev wrote:

if you look at your babelfish translation, it is not perfect, but we can surely all understand it



There's a huge difference between understanding the message and providing a final professional translation...

Personally, I would even feel embarrassed to use Systran, Babelfish or a similar program in order to "translate," let's say, a letter I originally wrote in English into Italian, French, German, and any other language I do not speak.

If you apply it to our professional reality, then I think it only makes things worse. Let's suppose that someone uses it to "translate" a resume or a cover letter and starts using it to contact clients. Do you really think someone will be able to attract more business that way? I'm sure the potential client will get the "gist" of your letter, understanding your intentions and general points howsoever, but I don't believe it's something that should be brought into the professional world of translations.

I also believe that the idea of having such program "translate" the material to get the hard part out of the way and then "correcting" everything to polish the final version is equally absurd. We still need human beings to differentiate the nuances.

Until languages "evolve" (?) to less complicated "codes" and words can be translated independently by a computer without resulting in any misunderstanding, humans with a lot of experience are still crucial to this process.


Direct link Reply with quote
 

Andrey Lipattsev  Identity Verified
Ireland
Local time: 00:53
English to Russian
+ ...
TOPIC STARTER
no, no, no? Oct 16, 2006

"I also believe that the idea of having such program "translate" the material to get the hard part out of the way and then "correcting" everything to polish the final version is equally absurd. We still need human beings to differentiate the nuances."

Thanks a lot for this input Rafa. This is, indeed, the kind of debate that I was looking for. However, allow me to make several points:

First of all, I did not quite understand your answer: it seems that you do not see machine translators as either an aid or as a competitor. Perhaps, you simply haven't seen a good one at work yet. I suppose that's what people thought about chess computers as they fisrt emerged, now the world champion as he is about to play Deep Fritz 10 says that his (the human's) chances are next to nill, but he will put up a fight for the sake of humanity. Chess and languages are different, and yet both are based on a set of clearly defined rules and logic. Linguistics is all about that.
Then again, there is poetry (in chess too), which is why the man can still beat the machine.

Secondly, I would like to know why you find "polishing" so absurd. In the context of using CAT (see my original message) I believe that this is what we do a lot of the time using these computerised tools. Perhaps, I did not express myself clearly enough at first. Of course, at the moment no program is able to produce a quality text. It lacks "poetry" and even more basically, it makes mistakes which it doesn't know how to spot or correct. There are too many variables for a software to take into account. I did not suggest that Systran (and I repeat Systran is NOT babelfish - the latter is the baby of the former) could at the moment replace a human translator. But that is exactly my point - I find it a much better aid than CAT tools, such as Trados. I also think that software and hardware development will make such programs better and better very quickly, which is why we do not need to wait for languages to "evolve" into codes (in my opinion, they are and have always been just that, but's that a topic for another thread), let's say not "codes" but "simplified codes", in order to see machine translations become efficient and accurate. I feel that we have to learn how to collaborate with these tools, how to use them progressively instead of rejecting them out of hand.

awaiting more rotten tomatoes...

[Редактировалось 2006-10-16 21:55]


Direct link Reply with quote
 

Riccardo Schiaffino  Identity Verified
United States
Local time: 16:53
Member (2003)
English to Italian
+ ...
Systran translations Oct 17, 2006

Andrey Lipattsev wrote:

What do you people think? Is Systran a replacement for us, humanoid translators, to be feared or is it a powerful tool that allows us to do our job quicker and better?


Systran (the company) translates its documentation in a number of languages.

And they don't use Systran (the tool) to translate their own documentation. Not even to pretranslate it (AFAIK).

Their translations are handled by human translators (aided of course by CAT tools).

So the answer to your first question is "no way": Systran may be useful enough for gisting, but it is still very far away from being our competition.

In certain circumstances, MT may become a useful tool for human translators (Jeff Allen has written various posts and articles about the uses of MT).

I think that we should also consider that at the beginning of IA (back in the 50s), MT was very poor and chess-playing programs (another mainstay of AI) were regularly beaten by mid-level human players.

Fifty years later, you can buy a Grandmaster-strength chess-playing program for just a few bucks, and the strongest programs have already beaten the strongest world champion ever (Kasparov). At the same time, MT programs are probably not stronger than before (the small gains they made could probably be ascribed to better and faster hardware).


Direct link Reply with quote
 

Rafa Lombardino
United States
Local time: 15:53
Member (2005)
English to Portuguese
+ ...
10x no Oct 17, 2006

Andrey Lipattsev wrote:

I suppose that's what people thought about chess computers as they fisrt emerged (...). Chess and languages are different, and yet both are based on a set of clearly defined rules and logic. Linguistics is all about that.


Yes, I agree with you that chess and languages are way too different. Chess works with coordinates (combining rows, lines, pieces, and allowed movements) and a somewhat limited set of rules when compared to languages, which are constantly evolving in a very complicated way, having words acquire new meanings everyday.

I still remember my first experience with computer-generated translations... I could not believe that the title "Miss" was translated as the verb "to miss" even though it was capitalized! C'mon, these are two very basic words that you can find in any elementary school student's dictionary! Having a background in programming, I was amazed by the fact that those (ir)responsible for such computer translator didn't even have the decency of inserting a code to differentiate a capitalized letter from its lower case "sibling."

Secondly, I would like to know why you find "polishing" so absurd. In the context of using CAT (see my original message) I believe that this is what we do a lot of the time using these computerised tools. (...) But that is exactly my point - I find it a much better aid than CAT tools, such as Trados.


Polishing a one-dimensional and completely blind translation provided by a computer is absurd, highly a big waste of time. I compare it to giving a text to someone who has no knowledge of the target language —let's say, the elementary school student— who'd try to translate it with the aid of a pocket dictionary only.

Not so long ago I actually had to proofread the English "translation" that a Spanish-speaking end client had worked on for their website. Needless to say, they had to pay almost twice as much for the proofreading task, considering the hourly rate, than they would have paid for the translation, to which we apply a rate per source word... They learned the lesson though, and since then they have been submitting the Spanish text for us to translate into English. I mean, you don't want to lose clients because your website sounds way too childish and sometimes even offensive to one market segment, right?

You can ask anybody if they'd rather translate the document themselves, considering of course that they do work with that language pair, or proofread a text translated by a machine or a non-speaker of the target language. If translators were merely "polishers" our job would be so "easy"... Have the client translate it by him/herself and then we just give that "extra touch", uh? Let's live a little more dangerously and completely disregard the source text, deciding what is correct by simply looking at the target text... Hum, that would be awesome! ... NOT! LOL

Anyway, polishing a multi-dimensional translation in which you or another competent linguist worked on with the aid of a CAT tool such as Trados is great and saves you lots of time. I'm not sure how familiar you are with such programs —and, for that matter, whether you can see a distinction between how CAT tools and computer translators work on a simply programming level—, but all CAT tools do is record the translation a human types and, in some cases, either allow you to compile your own dictionary or extract terminology from the translated text.

Once again, this is a text that a competent human translates. So the CAT tool uses code comparison to tell you that "the sky is blue" is pretty similar to "the sky is blue and clear," providing you with the previously saved translation you entered for the former and helping you formulate or complete a translation for the latter, maybe even showing you that in another project you determined that "clear" can mean X, Y, or Z in given contexts. Still, it's the human's call, never the machine's.

On another note, this discussion also reminds me of the poor level of literacy of a certain prominent politician, who while in college was trying to find a synonym for the noun "tears" because he had used it many times on his essay when talking about his little sister crying all the time, so he looked it up and found the verb "lacerates" and decided to use it instead and maybe score a big A+ for sounding so intelligent.

[Edited at 2006-10-17 04:57]


Direct link Reply with quote
 

Nadia-Anastasia Fahmi  Identity Verified
Greece
Local time: 01:53
English to Greek
+ ...
I am really.... Oct 17, 2006

Surprised that one is not able to make the distinction between a CAT tool and a computer translator. The concept is sooooo different.

And that is the key point in this discussion, as Rafa said before.

Perhaps, if pressed for time, I would deign to use Systran to translate a word - mind you "a word" - meaning I would deign to use it as a "dictionary" or "glossary" (although I would most probably have to make sure that the result is correct by checking in any other online glossary/dictionary - conclusion: waste of time).

Furthermore, your "translating" method seems to me such a waste of time, especially using Systran. Compared to the Trados, let us say "Translate" function, Systran is so elementary. Your method, I have to agree again with Rafa, reminds me of the nightmare of editing a translation by a non-professional and - to top that - a non-speaker of the target language (and I've done my share of such jobs until I decided that I will simply no longer accept such jobs... simply grotesque!!).

I have been using Trados (and DVX, although Trados is my favourite, maybe because I use it more) for 8 years now, and I am still wondering how I managed without CAT tools before (and that's a long 20 years before...)

But, then... one man's breath is another man's death... who am I to judge which method is better? If it works for you... then continue doing so...


Direct link Reply with quote
 

Steven Sidore  Identity Verified
Germany
Local time: 00:53
Member (2003)
German to English
There is a better way... Oct 17, 2006

Andrey Lipattsev wrote:
Traditional CAT just slows me down most of the time - some clients like Trados, some SDLx, some DejaVu, etc. I can't really spend time to learn how to use all these tools to their full extent.


This says all I need to know about your line of reasoning.

It's possible to sit inside a car with the door open and kind of push yourself along with your leg, but taking the time to learn how to turn it on and drive it is usually worth the effort if you want to travel long distances.

Systran is, a suppose, more like loading yourself in a catapult and firing away--you'll get there, but what's left won't look too pretty.


Direct link Reply with quote
 

Jabberwock  Identity Verified
Poland
Local time: 00:53
Member (2004)
English to Polish
Not so different... Oct 17, 2006

Nadia Fahmi wrote:

Surprised that one is not able to make the distinction between a CAT tool and a computer translator. The concept is sooooo different.



Actually, it is not. All those who have received texts with notes: "These are 100% matches, do not touch them", might already feel that their work has been "machinized". The fact that the segments were once done by a human is of secondary importance - if computer translates texts arbitrarily applying segments out of their context, without any human intervention, then it is not that far from MT at all.

I suppose that in the future (not very near, I guess), the two approaches will practically merge, i.e. MT will translate segments based on the big (and I mean really big) bilingual corpuses, using similar sentences or even paragraphs instead of simple "word for word" translation. On the other hand, the CATs will be more flexible concerning the differences between "segments" and "glossary entries", so that they will be able to propose a sensible suggestion even if the segment as a whole has not been translated before.

Of course, the software will always make mistakes. However, humans make mistakes, too. When the best MT software will reach the level of the worst human translator, it will still be considerably cheaper to use, so it may become a viable alternative for those really low-end jobs. Therefore, if you are really good, then you have nothing to be concerned about. If your only advantage is that rock-bottom rate, start worrying!


Direct link Reply with quote
 

Andrey Lipattsev  Identity Verified
Ireland
Local time: 00:53
English to Russian
+ ...
TOPIC STARTER
hmmm... Oct 17, 2006

I am a little surpised to see how personal this thread has become. I must have stepped on some pretty sore blisters. I am sorry if I did.
I am glad that people seem to understand my mindset so well. That always facilitates conversation. However, I would prefer if I could get more response to the actual issues that I am trying to analyze though this discussion and not just remarks about me and my work being catapulted somewhere.
Nadia wrote very interestingly that if my method works for me then I should continue with it. Thank you, Nadia. Not only does it work for me, I can quite honestly state that since I have dumped Trados and started using Systran my productivity rose, my quality has improved and I haven't had a single client complaining about the end result. Perhaps, sometimes it is better to get out of a VW Beetle that you've been pushing with your legs and get into a modern vehicle.
But I don't want to push the metaphor too far.
I suppose a key issue that seems to be coming to the fore in this discussion is that different methods work for different people. But that is something fairly obvious and I wouldn't start a thread just to discuss it. What I am a lot more interested in (especially now after all this feedback) is:
1. Why do you find MT (thanks for the acronym, Riccardo!) less useful than CAT? Rafa made some very good points here, but yet again it seems to me like she has never seen a good MT program in action. In all honesty, how often do phrases like "the sky is blue and clear" pop up in a text on a regular basis? Before I started to use Systran I have used Trados for a year or so. I was very excited with this program when I first saw what it did and how. However, with time I became quite disillusioned. I would have texts of 50K-60K words, sometimes over 100K, scientific texts, medical ones and all that Trados did for me was slow me down because of having to constantly press alt+ to get to the next sentence. The rate of repetition in a large free flowing text is next to 0 (in my experience). Perhaps, this should lead us to the next question:
2. Are there certain specific areas of translation where CAT tools are particulalry useful? Should we try to establish what these are, so as not waste our time on others where CAT does not cut down on the time we spend nor does it improve qiality?
3. And, finally, why can we remain convinced that MT will never reach the level of human translation? Again, thanks to Rafa for some very poignant remarks. I would, however, disagree with breaking the parallel to chess. The rules of languages are complex but definable. If up until now the programmers have been "irresponsible" that's rather their fault than a death sentence to MT in general, is it not? I would like to thank Riccardo for his comments on this. Could you, possibly, provide us with some links to these works by Jeff Allen that you refer to? Yet I would postulate that the lack of progress in MT over the past 50 years, as compared to computerised chess, has not been caused by the impossibility of such progress, but rather by it's relative difficulty and lack of real interest on the part of the community. In fact, as even this limited discussion shows, there are some deeply rooted interests at stake here and, obviously, whereas chess computers have been pushed forward by chess enthusiasts, MT is being, if anything, held back by professional translators. Isn't that an interesting phenomenon?
I must say, though, that I am extremely grateful for all the response that this thread has been getting. Thank you, everyone! And I hope we can keep it going for a while longer.

I have just read the post by Jabberwock and would like to thank him for such a positive and interesting contribution. I very mush share the opinion that this is the way, in which we are headed and we should take steps to keep abreast of the latest developments in both MT and CAT.

[Редактировалось 2006-10-17 15:29]


Direct link Reply with quote
 

Andrea Appel
Canada
Local time: 18:53
English to German
+ ...
I had to see for myself Oct 17, 2006

I did a copy and paste into systran and the same text into babelfish. Both the same thing. Of course I am not sure if this is only the website try out version and if the program is getting double as good once you buy the licence. If this is the case than please e-mail me instantly.

I am really against machine translation as well as CAT Tools, I also agree that it slows you down and the home page of ProZ.com is evidence enough; what should I do my CAT Tool did this and did that. I would hate to see that a translator comes back telling me that he can not deliver because of a problem with a CAT Tool.

It is kind of enticing to consider unrevised machine translation for certain customers, I mean the 0.03 cent ones. I also find that it is a big problem to get a source text which is machine translated and the customer wants it to be translated into a different language, sometimes not even aware that his copy is a machine translation. So than we have to bad translations. No no no I dont agree with any of this. Okay got to go and check out Systran. (-;

Oh and I do not intend to offend anybody/
We are all different so if it works for some people why not. I kind of see it as a bonus for them and for me. Me because I can offer human translation and them because they probably make better money with less effort.

I think with all of those tools we cut into our own fingers in the long run!

Bye bye all
Andrea

[Edited at 2006-10-17 13:41]


Direct link Reply with quote
 

Riccardo Schiaffino  Identity Verified
United States
Local time: 16:53
Member (2003)
English to Italian
+ ...
Real reason of Chess playing program with respect to MT Oct 17, 2006

Andrey Lipattsev wrote:

Could you, possibly, provide us with some links to these works by Jeff Allen that you refer to?


I don't have a link, but it should be possible to find them by doing a search on the forums: he regularly contributes to discussions about MT.


Yet I would postulate that the lack of progress in MT over the past 50 years, as compared to computerised chess, has not been caused by the impossibility of such progress, but rather by it's relative difficulty and lack of real interest on the part of the community. In fact, as even this limited discussion shows, there are some deeply rooted interests at stake here and, obviously, whereas chess computers have been pushed forward by chess enthusiasts, MT is being, if anything, held back by professional translators.


I'm afraid you are out on a limb here - the chess community has had really little to do with the progress of chess-playing programs: it has been pushed by programmers traditionally interested in this kind of problem. In any case the progress of chess as against the lack of progress in MT has to do with the different approach that was eventually adopted: it started out as a real AI project, trying to mimick the way humans play chess, and did not progress very far that way. The breakthrough came when they started to use brute strength approaches: a computer can calculate millions or even billions of different positions (whereas human grandmasters concentrate on the lines they "know" are most promising).

But such a "brute strength" approach to MT translation would be impossible or much more difficult: chess has a limited amount of rules, pieces and squares, whereas there are many more (and often ambigous) rules in human language, and I think that that is, essentially, while generic MT has not gone far.

Perhaps there is an approach to MT that will work in the long run: perhaps it is corpora (or statistical) based MT - such as has been used by Language Weaver, for instance. If so, it will be an approach closer to CAT than to traditional rules-based MT such as systran.


Direct link Reply with quote
 

Riccardo Schiaffino  Identity Verified
United States
Local time: 16:53
Member (2003)
English to Italian
+ ...
CAT tools Oct 17, 2006

CAT tools slow down the translators who do not know how to use them.

For the rest of us, they work fine (even with their bugs... but is there any complex software that does not have any bug?).

[Edited at 2006-10-17 19:31]


Direct link Reply with quote
 

Andrea Appel
Canada
Local time: 18:53
English to German
+ ...
Ohhh this means I dont know how to use it. Oct 17, 2006

Riccardo Schiaffino wrote:

CAT tools slow down the translators that do not know how to use them.

For the rest of us, they work fine (even with their bugs... but is there any complex software that does not have any bug?).


Direct link Reply with quote
 
Pages in topic:   [1 2 3] >


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


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

Do you see Systran as an aid or as competition?

Advanced search







PDF Translation - the Easy Way
TransPDF converts your PDFs to XLIFF ready for professional translation.

TransPDF converts your PDFs to XLIFF ready for professional translation. It also puts your translations back into the PDF to make new PDFs. Quicker and more accurate than hand-editing PDF. Includes free use of Infix PDF Editor with your translated PDFs.

More info »
Across v6.3
Translation Toolkit and Sales Potential under One Roof

Apart from features that enable you to translate more efficiently, the new Across Translator Edition v6.3 comprises your crossMarket membership. The new online network for Across users assists you in exploring new sales potential and generating revenue.

More info »



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