Wednesday, November 18, 2015

Translator Perspectives on MT & Technology In General

I found an interesting series of blog posts by Christelle Maignon that I thought articulated translator perspectives on MT and the increasing use of technology in translation work very well. She herself was driven away from translation work towards coaching because PEMT was just not her cup of tea from what I could gather. Anyway I thought it would be good to highlight her work in case you are not aware of her blog.

Some posts that readers of this blog may also find interesting are listed below:

Why machine translation creates so much anger and how to deal with it

This post references Dr Kübler-Ross study of grief. She describes the five stages of emotions which are experienced by people who are approaching death or dealing with the death of a loved one. Her model was widely accepted and it was found to be valid for other forms of losses, as well as situations relating to change (for instance, the loss of a job or of a familiar way of doing things). Her model has been used as a change management tool by businesses across the world.

I have written about this as well in the past referencing this link  but it is good to get a real translators perspective which interestingly uses the death and grief cycle as a reference.

Disruptive Change graphic

Another post describes the widespread use of MT based on presentation by Stefan Gentz and is one the most popular posts on her blog.

What Does The Future Hold For Translators?

I find the reaction and interpretation by a translator interesting though I don’t really see how MT is taking work away from translators or the professional translation industry. MT mostly translates stuff that would never get translated were it not possible to do it with MT.

Another that I found interesting is:

Riding The Wave Of Technological Change As A Translator

Or Future Proofing Your Career As A Translator

I think there is lots of useful information for translators on her site, and while I am regularly reminded that I am not a translator and should not be telling translators how to do what they do, I will dare to say that many will find useful information here.

I truly hope that my highlighting her blog here raises her profile and does not have a negative reaction from some who might see this as an endorsement from MT advocates.

I have not been very active in the last few months but I have a new series of ideas that I will start writing on again shortly.

Have a wonderful Thanksgiving vacation for those of you who celebrate this.

Let what comes come.
Let what goes go.
Find out what remains.

—  Ramana Maharshi

Himalayan view


  1. Thank you very much for highlighting my posts in your blog. Much appreciated!

  2. Hello Kirti,

    The common theme of Christelle Maignon's blogs seem to be acceptance of something that she feels to be inevitable, a fatality. She apparently believes that she can be replaced by the machine translation, to such an extent that she decided to do something else.

    I however, remain unconvinced and would like to add the following line to your quote from Ramana Maharshi:

    "Let what comes come.
    Let what goes go.
    Find out what remains.

    (and my contribution) "Hold onto and defend what you have got".


    1. Charles

      Thank you for your comments. My intent in choosing the referenced quote was to find something with a Thanksgiving theme, but I think I missed how it could also be linked to the MT theme.


    2. Thank you for your comment, Charles. To clarify, I don't actually see machine translation as a fatality but I do accept that it is here to stay and that it will continue to improve. I haven't stopped translating, and the reason why I decided to become a coach is because it was another career I had been meaning to pursue for a while. When one of my biggest clients switched to PEMT, my initial reaction was to worry, and it gave me the push I needed to commit to that other career. It was a personal choice and since then I've realised that translators have a lot of options.

  3. Why machine translation creates so much anger ?

    first, the general translation quality is falling down to unacceptable levels, at the point I regularly deactivate automatic translation on every website, if I want understand something

    second, too many (poor and lazy) peers have a blind truth in Internet (that is more and more MT managed), so their translation quality is falling down: a unsettling catch-22 situation

    add to this the plethora of thiefs (don't buy software/books, etc), and you have the reason why respectful dictionaries are no more translated (in Italian at least) so reducing translator quality: a unsettling catch-22 situation

    is it not enough to be angry?


  4. About all the reference to anger... that's a strong feeling.
    I think it pays to frame the Quality argument in perspective of the function of the communication and type of audience. In other words, what is the service demand.
    To make a practical example, you would not look for perfect quality in the output of a human professional simultaneous interpreter: the function trumps the form, the communication is established and the user is happy. Those quick translated utterances don't offend us and allow us to go from A to B efficiently.
    Same for MT.
    If we work in a section where communication function and speed or communication availability trumps perfect quality, we should embrace it as the reality of facts (we would not want that simultaneous interpreters should hesitate to utter their translations in search for the perfect choice - fast and accurate is their game given their function) or realize that there are other sections where speed/availability is not the key function and we can work on the ultimate quality there. I think it pays to understand that a service exists only in function of its demand and user/buyer needs. If demand/users/buyers needs are changing, I feel that resisting is irrational.

  5. There is a long and well-consolidated tradition in our culture to consider technological development a threat. Just think of William Blake’s “dark satanic mills”. Nothing new in this. However, the Luddites did not only fought against the machine, they also fought for decent salaries and values they feared would be overruled by the machines. In other words, if this whole MT against HT is boiled down to business – because this is eventually what it is all about – the conditions are not satisfactory for translators at the moment. I am not against MT at all, and I don’t have that much hands on experience with PEMT. From what I have done so far I got the feeling that the price dumping following in the wake of MT is punishing me as a translator/editor. When I am asked to post edit a (bad) machine translation to a miserable word price, just because it is a machine translation that supposedly needs only a light restyling, but actually requires a lot of time - not to be perfect but to be decent - well, then my business is not healthy any longer. Being just myself I can’t restructure my business sending employees back on the street. Also, there is a limit to how much I can expand my business = crunching more words, after all the day has only 24 hours. But my electricity bills do not come with a discount, on the contrary! If a client can live with the fact, that Microsoft not only translates Kirti Vashee to Lotte Nørgård in my native language (I tried with this page), but also transforms him into a female, that’s fine with me. But I would like to receive a decent pay for bringing him back to his natural state. :-)

    1. So what you are telling us, Birgit, is that machine pseudo-translation offers an additional benefit that few anticipated? Low-cost transgendering.

    2. What I find interesting about this whole debate is the difference between technology led and translator led translation vendors. I've worked for both. For the technology led ones, the vision of what the technology provides is the be all and end all. The reality of the technology is that it's not quite as plug and play and fully integrated with the other complementary tools as is made out. This is where lots of IT customisation comes in to make systems integrate and try and work seamlessly together. On the other hand, the translator led vendors focus constantly on the overall quality of translation output. If they find MT helps provide quality output more quickly, then they use it but if it affects overall quality they don't. I see a definite positive role that MT can and does play for technical authoring content where the source content is in dry and consistent language. However for anything more complex or creative, I am yet to be convinced it is fit for purpose. How professional translators translate and the types of content they work on is bound to continue to evolve but I don't see the prospect of the extinction of our profession any time soon.

    3. Birgit

      Thanks for your response and as you point out, MT is very literal as can be seen by how my name was handled. I think you are right that many agencies use the technology as a way to justify lower rates, often unfairly.

      I would suggest that if a translator understands the quality of the MT, they can then determine a rate that would be fair if they would even consider the work, or simply refuse to do the work. Many if the LSPs that are "using" the technology do not understand that it is possible to get a significant amount of variation on the quality of MT output and tend to automatically pay an arbitrarily lower rate without understanding the specific output they will hand over to editors.

      Thus I think it is wise for translators to learn to rapidly assess some "ACTUAL" MT output for the PEMT project and then determine a floor rate for themselves.

      Thus, an LSP who is not willing to provide a sample of at least 100 actual segments form the MT system should be automatically excluded, and once you examine a sample of the MT output you can make a determination of what a fair rate is.

      PEMT needs trust to be viable and in any low-trust situation I think PEMT is questionable work. But then any work in a low-trust situation should be questionable.

      I describe some possibilities on how this assessment can be done in earlier posts (see most popular posts)or look at this one:

      But I guess as you point out, PEMT work should always be a Translator Beware situation until you properly understand the VERY SPECIFIC output you will be working with.


    4. Thanks for hints and tips, and I will definitely sit down and read your papers about PEMT.
      Seen from my point of view as a freelancer one of the big gaps in the ongoing MT debate is the distance between top and bottom, meaning the proportion between what is written about MT and what a freelancer actually puts their hands on. Working with a minor language makes this distance even bigger. At the top (only figuratively spoken) the MT industry has now for decades been claiming that within short all translation would be done by MT engines, and at the bottom of the supply chain a lot of translators fear unemployment as the worst case scenario, or a best case scenario where time is spent doing PEMT at miserable rates sacrificing skills, creativity and originality ¬– and at the end also the uniqueness of language and language quality. The more responsible part of the MT industry and the LSP’s, in collaboration with translators’ association could do a lot to fill out this gap. What we need is much more (practical demonstrations) about how to make the best of it and discover new opportunities for translators. Why does nobody talk about the huge potential for pre-editing texts for MT? No one better than translators know how to cut a text into a structure and sentences that will optimize the MT output. I suppose MT should also lead to more attention on terminology, at least in major companies. How do we prepare for that? I am sure there are many other possibilities than just PEMT for who would like to be part of the MT development. So less declarations and more condivision of practical insight. The whole language industry has a big responsibility for implementing good translators and their skills. For centuries translation has been a driving factor in the historical, technological and cultural development, and still today translation is everywhere in everyone’s life. Invisible, but it is there. So more spotlight on the human factors and how to make the best of them would be a big step ahead.
      I also feel the need to cry out my desperation over the ongoing abuse of the large scale automated Google or Microsoft translation. Navigating the web, even with advanced search criteria for my specific language, automatically translated pages pop up everywhere like a wolf in a sheep's clothing. The first 10-15 hits are now almost inevitably Google translated pages. Doing revisions of translations it happens still more often to see a ‘translated’ term borrowed from one of these pages. I can see the democratic aspect in these automatic large scale ‘translations’, they might in some fortunate case help someone to get an idea of the content. But mostly the world would be a better place without them. What I am trying to say is that Google & Co. translations are not a good business card for serious MT translation. Somebody should start to explain and show the difference, and providers with automatic translated pages should have the guts, or even better be obliged to display a big warning sign on their pages. Why hide? Nothing to be ashamed of, but don’t sell it for what it is not, namely human language.

    5. Birgit

      Indeed over zealous MT advocates have a great proclivity to wrongly forecast the end of translation work. They remain quite far from this being true even though MT does continue to improve, especially in the hands of skilled practitioners. But even in 2015 the expert systems remain the minority and we see that the bulk of the bulk of the MT activity in the professional sphere is still incompetent Moses or Instant MT from Do-it-yourself services that most editors find hard to work with.

      These sub-par systems will only die away if more translators refuse to work with them. All PEMT work is not wretched, but a lot of it in 2015 is, and it would be wise for translators and editors who consider this kind of work, to learn how to tell a "good MT system" from the hack and sub-par attempts.

      "What we need is much more (practical demonstrations) about how to make the best of it and discover new opportunities for translators. Why does nobody talk about the huge potential for pre-editing texts for MT?" I completely agree with you on this as there is much that can be done to improve the experience by doing this.

      I also feel that MT will have only come to it's true promise and potential when professional translators want to use it because it makes their work easier and more productive. The MT systems and the user controls are evolving to a point that this can happen in the not so distant future.

      I think specific and articulate feedback like yours would be very valuable to developers who wish to produce useful tools that enhance professional translator productivity, and for the ever growing need for higher quality computer translation. I think the need for both is more aligned than not and work done for one will affect the other.

  6. Hello, Kirti and hello to my fellow translators. I got here because Jost Zetzsche tweeted about it.

    First, I do have PEMT experience, 10 solid months of well-paid MT postediting. And by “well-paid” I mean over $40/hour.

    Before you go all “WHAAAA!!!” on me, let me explain. I worked as part of a team of translators (several languages) for a cybersecurity software company in Silicon Valley (as a home-based translator). Our job was to translate a large amount of user documentation and build a glossary. We used Lingotek's online TM software, which siphoned MT-generated translations for us. We were told from the beginning that we, the translators, would have the last word on the translation.

    Some team members found little to edit and others, including me, found a lot to edit. Throw in differences in syntax and style from Korean to Spanish to German and you'll get a general idea. Some of the translators who found little to postedit in the MT segments were more “productive” than others. The company had little, if any, experience working with translators, let alone software localization specialists (which we all were) and translation or localization managers: to illustrate the latter point, our first manager was let go after 2-3 months into the project (which is ongoing, by the way). They changed managers 3 times during my tenure.

    I'm giving this background so that no one can call me a Luddite or an enemy of progress or technology. I like to deal with facts, not futuristic platitudes or prophecies.

    There is no way to future-proof any career. Why? Simply because we are human. Even the wisest among us cannot predict more than general circumstances and conditions 4-5 years out.

    While MT technologies, like all technologies, seem to be promising and offering more opportunities, its track record has left much to be desired. Now, why some translators are 'angry' at MT or PEMT? First, why use the word 'anger' in the first place? I've discussed PEMT with colleagues and they have shown impatience, unwillingness to discuss the topic, jadedness, even cynicism. But not anger. I see no need to either assume MT/PEMT opponents are angry (a very strong emotion) or to condescendingly project an emotion on them.

    Of course, PEMT is not everyone's cup of tea, just like movie subtitling is not for everyone. There is, I believe, plenty of variety in the translation world to keep most translators occupied, provided that they excel in writing abilities. You can be the most technologically equipped and PEMT-abled professional, but if your writing sucks (pardon my French), then you'll always be a mediocre professional, technology or no technology.