Pages in topic:   < [1 2 3 4] >
How do you use MT?
Thread poster: ..... (X)
Michael Beijer
Michael Beijer  Identity Verified
United Kingdom
Local time: 23:15
Member (2009)
Dutch to English
+ ...
@Dan + Samuel: Sep 2, 2015

Samuel Murray wrote:

Dan Lucas wrote:
There was a case widely reported in Japan early in 2015 in which it was found that people (including employees of government ministries and major banks) had entered various strings into an MT site based in China.


If you're talking about the article in The Japan News of February 2015, well, it was reported in only one newspaper (The Japan News) and then reposted verbatim on several blogs, but I was unable to find any other report about it in any newspaper in English. Do you happen to know the name of the MT system that they used? You seem to know that the MT system was based in China, but this fact is not reported in the The Japan News article, as far as I'm aware, so can you remember where you read that piece of information?

I would have really liked to double-check the journalist's comment that the users of that web site were unaware that their texts would be published on the internet.


As far as I know, Google Translate, e.g., does not store your source text or recycle it (unless you use their ‘Google Translator Toolkit’, and select certain sharing settings). Of course, I have no hard evidence that this is true. However, neither is there any hard evidence that what you say actually happens (apart from on obviously crap sites like www.ilovetranslation.com etc.).

Michael

[Edited at 2015-09-02 15:00 GMT]


 
Dan Lucas
Dan Lucas  Identity Verified
United Kingdom
Local time: 23:15
Member (2014)
Japanese to English
No other English source Sep 2, 2015

Samuel Murray wrote:
I would have really liked to double-check the journalist's comment that the users of that web site were unaware that their texts would be published on the internet.

Did they say that? To be honest, if the people using the service had known they might expose themselves, I don't think there's the slightest chance they would have pressed the button. Japanese in general and people in prestigious jobs in particular are hyper-sensitive to risk. I don't think the average Japanese person has any idea that MT might entail problems of confidentiality.

I clipped an English version of the article about the time it was released, which mentions that the Yomiuri phoned a Chinese company in Chongqing. However, I can't find anything else in English. There's nothing on the MT system in the article I have. The site, according to this post (with screenshots) was ilovetranslation.com.

The original article is reported to have appeared in the Yomiuri Shimbun on 20 February 2015, on the front page. I can't confirm that directly but the source appears to be a Japan-based security company called Lac, which confirms that one of its executives was quoted in the Yomiuri article appeared but gives no detail.

The story also appeared in the Sankei Shimbun in this article on the same day. Also in the Asahi Shimbun here, although truncated for non-subscribers.

Regards
Dan


 
Heinrich Pesch
Heinrich Pesch  Identity Verified
Finland
Local time: 01:15
Member (2003)
Finnish to German
+ ...
Often surprisingly good Sep 2, 2015

I use Google MT if I'm allowed to and sometimes I'm surprised how good it works, but of course often it fails. For me MT is a modern version of the dictionary, only that MT often finds terms that are not in any dictionary. It would be more useful with a few adjustments. So it should write "a/b" and not "a / b" and some other stuff that does not come to mind now. Probably MT will never grasp German grammar completely.

 
Samuel Murray
Samuel Murray  Identity Verified
Netherlands
Local time: 00:15
Member (2006)
English to Afrikaans
+ ...
@Dan: Hmm, option to share is checked by default Sep 2, 2015

Dan Lucas wrote:
The site, according to this post (with screenshots) was ilovetranslation.com.


Well, currently (and in the screenshots of that post, if I see it right), the option to share the translation is checked by default.

I only saw that option after I did my first translation -- before that, my eyes ignored it along with all the other ignorable content on the site. That's the problem with web sites these days -- there's so much junk on it that you learn to see only what you think you need to see.



 
Bernhard Sulzer
Bernhard Sulzer  Identity Verified
United States
Local time: 19:15
English to German
+ ...
Not a pretty thing Sep 2, 2015

Kevin Dias wrote:

Hello everyone,

I'm interested in learning more about how (or if) translators use machine translation in their translation workflow.

- Do you use machine translation?quote]


I think it would be a good idea to learn your definition of MT. We have been discussing this topic for a while now here on Proz.com. Plenty of threads. I am afraid that people mean different things when they talk about MT and the "benefits" involved. Throw in the new "post-editing MT" craze promoted by interested parties, and it seems oh so natural to "use MT" in your work process, specifically by "simply" "editing" the text a "machine" "translated."

Word of warning: "Translating" is something humans do. Using bilingual segments such as translation memories is not "translation" and it's not "machine translation."

I don't use texts that such machines have "produced" and then start "editing" them, expecting it to be a task comparable to editing the text of a professional or even amateur human translator.
That's an important point because this is what some outsourcers are trying to do: get a language salad fixed at the price of an editing job - well we're talking about unprofessional outsourcers, to be clear, of which plenty can be found on any translation portal.

And what about "training translation engines??" Personal translation engines? If you are referring to bilingual segments you added into a machine (what machine please?!), and later use it for new texts, that's not "translation" either. It's a TM that you might be able to trust because you created it and you expand it, but it's up to the human translator (the only kind of translator) to decide what to use. But that would be terminology and certain phrasings, but in no way a text that your "machine" produced by itself, ready for editing. And I am talking about "text," not simply a list of words.

Even less of a trustworthy source, even if you argue for it as a help file, would be the text an agency sends you that they have run through a translation machine.

Machine "translation" is a misnomer. machines don't translate and thus I have 0% trust in it.

As in: "We have an MT text that needs to be "edited" by tomorrow (6000 words), please quote your best editing rate." That's what people out there are trying to do. Shameful!

If MT is not "translation" (and it isn't), whatever it is cannot be the basis for an editing job. Poof.

So before we nonchalantly talk about MT as if it's such a sensible concept or ask do you use it, we should examine the basic facts involved. They're not pretty AFAIC.

[Edited at 2015-09-02 16:01 GMT]


 
Michael Davies
Michael Davies  Identity Verified
Denmark
Local time: 00:15
Member (2009)
Danish to English
+ ...
- Do you use machine translation? Sep 4, 2015

I use MT unless my client specifically forbids its use.

I find that it saves me a lot of keystrokes (and, not least important, time!) during the translation process.

I find that it shifts the work load from translating texts to proof-reading as I always use 2 cycles of proof-reading during the translation process:

  • in the first cycle, I do a quick run through the translated text, correcting the worst errors (and it is only exceptionally that there are TU... See more
  • I use MT unless my client specifically forbids its use.

    I find that it saves me a lot of keystrokes (and, not least important, time!) during the translation process.

    I find that it shifts the work load from translating texts to proof-reading as I always use 2 cycles of proof-reading during the translation process:

  • in the first cycle, I do a quick run through the translated text, correcting the worst errors (and it is only exceptionally that there are TU's, which do not require correction) and making sure that the meaning of the translated text corresponds (more or less) to the meaning of the source text.


  • after completion of the first cycle, I perform an quality check of the translated text (using my CAT's verification function to catch spelling and grammatical errors, before I make a final check of the complete translated text to make sure that it reads naturally and there is then no difference between the final result and the result of a manual translation.


    This may seem to be an arduous process but I have timed the duration of translation of texts, which are similar in volume and there is (usually) a time saving using the MT / post-editing method. However, if the source text is of poor quality (maybe a poor translation from another language) then it does require more work than if it is of good quality since the MT struggles to get it right.

    I use mainly MyMemory and SDL Language Cloud and, more occasionally, Microsoft's MT Enhanced using Microsoft Translator (particularly for texts within the IT domain) and sometimes Google Translate (which, in my view, has vastly improved since I first used it a few years ago). ▲ Collapse


  •  
    Michael Beijer
    Michael Beijer  Identity Verified
    United Kingdom
    Local time: 23:15
    Member (2009)
    Dutch to English
    + ...
    can you give us a bit more info on how Microsoft's "MT Enhanced" works? Sep 4, 2015

    Google Translate Microsoft Translator Google Translate
    Michael J.H. Davies wrote:

    I use MT unless my client specifically forbids its use.

    I find that it saves me a lot of keystrokes (and, not least important, time!) during the translation process.

    I find that it shifts the work load from translating texts to proof-reading as I always use 2 cycles of proof-reading during the translation process:

  • in the first cycle, I do a quick run through the translated text, correcting the worst errors (and it is only exceptionally that there are TU's, which do not require correction) and making sure that the meaning of the translated text corresponds (more or less) to the meaning of the source text.


  • after completion of the first cycle, I perform an quality check of the translated text (using my CAT's verification function to catch spelling and grammatical errors, before I make a final check of the complete translated text to make sure that it reads naturally and there is then no difference between the final result and the result of a manual translation.


    This may seem to be an arduous process but I have timed the duration of translation of texts, which are similar in volume and there is (usually) a time saving using the MT / post-editing method. However, if the source text is of poor quality (maybe a poor translation from another language) then it does require more work than if it is of good quality since the MT struggles to get it right.

    I use mainly MyMemory and SDL Language Cloud and, more occasionally, Microsoft's MT Enhanced using Microsoft Translator (particularly for texts within the IT domain) and sometimes Google Translate (which, in my view, has vastly improved since I first used it a few years ago).


  • Interesting Michael, can you give us a bit more info on how this works? I keep meaning to check out the Microsoft Translator Hub. I use Google Translate and Microsoft Translator (via their respective APIs), both in little boxes in my CAT tool, and sometimes my CAT tools's various little tricks where it tries to combine them with my own resources, and auto-assembly, etc.

    Michael


     
    Jeff Allen
    Jeff Allen  Identity Verified
    France
    Local time: 00:15
    Multiplelanguages
    + ...
    Interview with Philipp Koehn about Statistical MT Sep 5, 2015

    Meta Arkadia wrote:
    I think the current common approach (statistical) is wrong and its evolution too slow, and although it may lead to acceptable results in the near future, it can never be better than "acceptable."


    Hans,

    In June 2015, I was an invited panelist to ask questions in: An Interview with Professor Philipp Koehn - MT Past, Present and Future (sponsored by AsiaOnline).
    Philipp is the creator of the Moses decoder and wrote the book "Statistical Machine Translation". Some of the questions and answers focused on where MT is going, especially with regard to rule-based systems, example-based, and now stat-based.
    Here is the link to download the webinar:
    http://www.asiaonline.net/EN/Resources/Webinars/#Webinars21

    Jeff


     
    Jeff Allen
    Jeff Allen  Identity Verified
    France
    Local time: 00:15
    Multiplelanguages
    + ...
    There are non-pivotal (English) MT systems/software Sep 5, 2015

    Andrea Halbritter wrote:
    I do not use it at all. In my language pair the result is not understandable.


    Meta Arkadia wrote:
  • As far as I know, ALL of them use English as a "conversion"/"bridge" language, in other words, your original French text is translated into English, and the resulting English text is translated into German (which is why you'll find English words in the German text). problems? Here's how to get them


  • There have been MT software programs/systems which do not use English as a pivot.

    The PROMT-based Reverso (v3, v4, v5) had direct French German processing. I have licenses for those and used them for years. The problem now is a compatible operating system. I'm setting up Virtual Machines at the office for running these translation systems on Win XP and earlier OS versions.

    Jeff


     
    Jeff Allen
    Jeff Allen  Identity Verified
    France
    Local time: 00:15
    Multiplelanguages
    + ...
    be careful with contracts stipulating no use of MT Sep 5, 2015

    Dan Lucas wrote:
    However. With most of my clients I have signed conditions prohibiting me from using MT. And even with their permission, I would be very wary of using anything that passed sensitive information - such as, for example, a financial report that has not yet been released - back to the MT provider.


    Signing these types of agreements show that the customers are very ignorant regarding the technology types. I have said for 20 years that Translation memory is just the commerically interesting and sexy name for Example-based Machine Translation.

    The different types of Computer Generated Translation
    https://www.youtube.com/watch?v=Po0mEXwl9Dk

    And in round table discussions and articles in which I invited many of the early TM/CAT tool providers to participate while I was the moderator or issue editor, they have admitted that ther system was based on the principle of Example-based Machine Translation (and even said that it was EBMT).
    This means that if you are using any type of TM / CAT tool that provides segment-level, memory-based translation processing, then using those solutions could also be considering as violating such agreements.

    Jeff


     
    Michael Beijer
    Michael Beijer  Identity Verified
    United Kingdom
    Local time: 23:15
    Member (2009)
    Dutch to English
    + ...
    Confidentiality issues, rather than type of MT system Sep 5, 2015

    Jeff Allen wrote:

    Dan Lucas wrote:
    However. With most of my clients I have signed conditions prohibiting me from using MT. And even with their permission, I would be very wary of using anything that passed sensitive information - such as, for example, a financial report that has not yet been released - back to the MT provider.


    Signing these types of agreements show that the customers are very ignorant regarding the technology types. I have said for 20 years that Translation memory is just the commerically interesting and sexy name for Example-based Machine Translation.

    The different types of Computer Generated Translation
    https://www.youtube.com/watch?v=Po0mEXwl9Dk

    And in round table discussions and articles in which I invited many of the early TM/CAT tool providers to participate while I was the moderator or issue editor, they have admitted that ther system was based on the principle of Example-based Machine Translation (and even said that it was EBMT).
    This means that if you are using any type of TM / CAT tool that provides segment-level, memory-based translation processing, then using those solutions could also be considering as violating such agreements.

    Jeff


    Good point. However, I think what they are really worried about when they say "No MT!" is not the type of MT engine system involved, but rather, the confidentiality issues supposedly surrounding using Google Translate or Microsoft Translator plugins in your CAT tool. These people seem to think Google saves and might leak their source texts if you use MT plugins in your CAT tool (which isn’t the case, as far as I can tell).

    Michael


     
    Meta Arkadia
    Meta Arkadia
    Local time: 06:15
    English to Indonesian
    + ...
    Confusion Sep 5, 2015

    Jeff Allen wrote:
    There have been MT software programs/systems which do not use English as a pivot.


    I know. But most people - including some of the posters on this thread - seem to think MT=GT. That there's a lot going on outside Mountain View, escapes them. And I was referring to

    But I think you're referring to the MTs you can consult on/via the Internet


    and for most people that's limited to The Thieves of Mountain View, Bing, MyMemory. In that order. The free or for peanuts MTs you can consult and integrate in your CAT tool. The OP clearly isn't interested in those machines.

    Cheers,

    Hans


     
    Meta Arkadia
    Meta Arkadia
    Local time: 06:15
    English to Indonesian
    + ...
    There goes... Sep 5, 2015



    ... my Sunday. Thanks anyway.

    Cheers,

    Hans


     
    Jeff Allen
    Jeff Allen  Identity Verified
    France
    Local time: 00:15
    Multiplelanguages
    + ...
    Translation memories are in fact one type of MT Sep 5, 2015

    Bernhard Sulzer wrote:
    Word of warning: "Translating" is something humans do. Using bilingual segments such as translation memories is not "translation" and it's not "machine translation."


    Bernhard,

    Translation memories are in fact one type of MT. I've just mentioned this in my immediately previous post in this thread. Over the past 15-20 years, I've given probably 2 dozen conference talks on this subject (many of them even being peer reviewed articles), and often in the presence of colleagues from the various TM/CAT tool solution providers. Even gave a special invited talk to the Trados development team in 1998.
    Not a single person from any of those companies, nor any review comments received for the submitted articles, have ever mentioned that my statements were off-track from a technical standpoint nor from a marketing/sales perspective.
    On the contrary, a VP from SDL sitting next to me during the presentation at Localization World 2004 asked afterward for a copy of the presentation because it helped simplify the explanation of such technologies.

    The term "Translation Memory" likely continues to be used by all such solution providers because it is just a more user-friendly term than "Example-based Machine Translation".

    Andy Way is the only person who has told me that he thinks there are some differences. This was several years before he left the academic arena to spend 3 years in commercial translation software and service companies. I should follow up with him now and see what he thinks after having lived daily in such trenches for those 3 years.

    There have been dozens, even hundreds, of posts in these various discussion groups over the years in which freelance translators claim that TM is not MT. Yet, there is no substantial evidence, and that being based on technical descriptions, that TM is different from EBMT.

    Here is one written statement along these lines:
    ELRA Newsletter
    Vol.4 N°3, July-September 1999
    Page 5, column 2: "This is very similar to Example-Based Machine Translation."
    Article title: Beyond “fuzzy matching” – The Déjà Vu approach to reusing Languages Resources
    by Xavier Garcia
    http://www.elra.info/media/filer_public/2013/09/06/v4n3.pdf


    Jeff


     
    Jeff Allen
    Jeff Allen  Identity Verified
    France
    Local time: 00:15
    Multiplelanguages
    + ...
    yes, more related to quality and confidentiality issues Sep 5, 2015

    Dan Lucas wrote:
    However. With most of my clients I have signed conditions prohibiting me from using MT. And even with their permission, I would be very wary of using anything that passed sensitive information - such as, for example, a financial report that has not yet been released - back to the MT provider.


    Jeff Allen wrote:
    Signing these types of agreements show that the customers are very ignorant regarding the technology types. I have said for 20 years that Translation memory is just the commerically interesting and sexy name for Example-based Machine Translation.
    ...
    This means that if you are using any type of TM / CAT tool that provides segment-level, memory-based translation processing, then using those solutions could also be considering as violating such agreements.


    Michael Beijer wrote:
    Good point. However, I think what they are really worried about when they say "No MT!" is not the type of MT engine system involved, but rather, the confidentiality issues supposedly surrounding using Google Translate or Microsoft Translator plugins in your CAT tool. These people seem to think Google saves and might leak their source texts if you use MT plugins in your CAT tool (which isn’t the case, as far as I can tell).


    And I totally agree with you on that Michael. My experience on this going back to the 1990s with such agreements has been that customers are worried on at least 2 points:

    1) that a translator/translation post-editor is not just sending the text through MT and then delivering the text as-is without additional revision/checking. As so many people have done this in the past, this has led to the contractual statement.

    2) the content confidentiality issue. This has become more important since approx 2006 when Google finally launched their own statistical MT version of Google Translate (replacing their license of the then rule-based SYSTRAN that was behind Google Translate up until that point).

    However, legal articles and clauses can be used in other ways, if and when necessary. I've had my share of participating in the analysis of a few legal cases in the area of translation technologies, especially where the claims went quite far.
    And then another similar type of situation emerged of the past few years over a patent war on certain types of these technologies. That one went quite far, and even to the point where the appeal did in fact reverse the decision. I cite that example in courses on software intellectual property best practices which I give at SAP to software development teams.

    Jeff


     
    Pages in topic:   < [1 2 3 4] >


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


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

    How do you use MT?






    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 »
    Trados Studio 2022 Freelance
    The leading translation software used by over 270,000 translators.

    Designed with your feedback in mind, Trados Studio 2022 delivers an unrivalled, powerful desktop and cloud solution, empowering you to work in the most efficient and cost-effective way.

    More info »