Posted by: patenttranslator | August 17, 2016

Is There a Market for Post-Editing of Machine Translation in Your Field?

The argument that “GoogleTranslate” translates more words per minute than all human translators in one year is much bandied around as a rationale for why human translators must adopt machine translation (MT) as a tool for their own work not only in PR materials (commercial propaganda) of “the translation industry”, but now also on some translators’ blogs, usually blogs of former translators working for “the translation industry”, or of those who use their blogs as a marketing tool to attract newbie translators to their paid webinars and other offerings. 

This argument is usually accompanied by statements like these:

  1. The market for human translation and for MT is worth X billion dollars already and it is growing by X percent a year.
  1. Translators and “the translation industry” have already lost 99% of our market opportunity to Google and Microsoft.
  1. We, human translators, must embrace technology instead of stubbornly resisting it and use it to work faster because that’s what clients want.
  1. Technology is progressing so quickly that MT will be “almost as good as human translation” within (fill in a number of years). MT will be just as good (or almost as good) as human translation very soon (although estimates vary, from three to 20 years from now; there has not been not much change in this respect in the last two decades or so).

Et cetera, and so on and so forth – this kind of philosophical prognostication and prophesying has been going on for quite a while now.

All of these arguments are somewhat valid, as far as they go, at least from the viewpoint of “the translation industry”, i.e. the people who buy translations from translators and sell them at a higher price to their clients. “The translation industry” would obviously love to be able to gorge itself on a bigger piece of the pie, especially since it is such a huuuge pie, as Bernie Sanders would put it. (I think that it is important to make the distinction between actual human translators and “the translation industry”; that is why I always use quotation marks when referring to this particular industry.)

Like many other translators, I have been using machine translations as a tool for my own purposes – mostly to give me an idea of the material that I am about to translate – for more than a decade.

But based on the experience of this human translator who has been trying to figure out whether I can use machine translation to supplement my own income, namely income derived from my own human translations and from the work of translators who kindly work for me in exchange for the money I pay them (which also makes me part of “the translation industry”), none of the arguments listed above really makes a whole lot of sense.

First of all, it is clear to me that nobody can put a number on the “value of the market for translation”, whether it is human translation or machine translation.

If one looks at the constantly growing global population of people who need strong headache medicine because their life is constant suffering and pain, the market for Percocet, a strong, combined opioid/non opioid pain reliever, would be certainly on the order of several billion people, and thus could be valued at many trillions of dollars. And although different pharmacies and stores charge different prices, according to an answer that I found on Yahoo, Percocet costs about 300 dollars (I assume this is per bottle, not per pill, at least not yet). Unfortunately, this means that 99% of the people who might want to use it because their life is constant agony, torture and pain cannot afford it. In third world countries, it would be more like 99.99% of people who are unable to purchase this wonderful drug.

The high price of Percocet and other opioid and non-opioid drugs is thus probably why most people simply opt for booze.

The estimate of the monetary value of the worldwide market for human translation is similarly nonsensical if 99.99% of people who might benefit from translations can’t afford to pay for them. Since human translation, especially human translation obtained from highly educated and highly experienced translators, is very expensive, even many people who could afford human translation decide to do without it and opt instead for machine translation, or no translation at all.

And even though what passes these days for human translation, namely the stuff “the translation industry” is selling to initially unsuspecting clients, is somewhat less expensive, it is still expensive and will remain so no matter how many trillion words “must be translated”, supposedly because they are simply there.

The high cost of human translation is also one reason why “the translation industry” has been telling us for quite a few years now that the market for machine translation is worth X billion dollars and that Google and Microsoft have already captured 99% of this market while both translators and “the translation industry” were asleep.

But there is a good reason why Google and Microsoft have quasi-monopolized this market: the MT service provided by Google and Microsoft is free, unless you need huge quantities of MT, in which case the cost is still quite miniscule.

So how do we as translators compete on price with a service that has been free already for more than a decade as “translation industry” entrepreneurs?

I remember that about 20 years ago, when machine translation that almost made sense, at least some of the time, was still a new-fangled invention, a patent lawyer who found my website online called me to inquire how much would I charge to edit a machine translation of a Japanese patent for him. He insisted that the machine translation was “pretty good”, but that it still needed some editing. I remember that he said the words “pretty good” two or three times.

I declined to help him because I did not want to downgrade the value of my services to such a low level at that point, and also because I did not know what else to say and how much to ask for.

Ever since then, or for about 20 years, I have been intermittently trying to figure out how to use MT to actually make money by editing it as a highly experienced, human translator, knowledgeable in the field of patent translation, given that I have been working in this field for almost three decades. Specifically, I have been trying to figure out how to make money from MT not by working as a slave for a “translation industry” intermediary, but on my own, when I work for a direct client.

As you have probably guessed by know, even though I would find this kind of work quite distasteful, I will do just about anything for money.

But although I must have offered this kind of service dozens of times, so far there have been no takers. I don’t offer post-editing of MT detritus very often. But if I feel that the client is very “price-sensitive” (cheap, looking for a bargain), in order to get my foot in the door, so to speak, I sometimes offer several options to a new potential client:

1) A full translation (option A), which might be for example a thousand dollars,

2) Translation of claims and the text describing figures, which might be for example a hundred and fifty dollars (option B),

3) Post-edited machine translation (option C) for the same price as option B, usually if it has been a slow month.

This sometimes results in takers for option B, but so far, there have been no takers for option C. The problem is, basically all of my clients and even potential clients not only know how to get a free machine translation of the text of a patent, but they also know that edited machine translation would be of such a low quality that it still will not really be worth the money that I am asking for it.

Which does not mean that there is no market there for edited (post-processed) crap, which will still be crap, although on an improved level – we could call it high-grade crap.

But since my experience is basically anecdotal, it may be applicable only to the relatively narrow field of patent translation. There must be some materials that do not really need to be translated, but that could make managers of some enterprises look very sophisticated, value-conscious and forward-looking and all that if they could show graphs and flowcharts and other props showing how much information they have been able to obtain with translations at a very low cost.

Translation of materials that do not really need to be translated, but that might make management look good if these materials were translated, is probably the prime market for post-edited machine translations.

But is there a market for post-edited translations of patent applications? My experience so far seems to indicate that the market for post-edited translations of patent applications does not exist because these post-edited translations would need to be highly accurate.

For better or worse, there just does not seem to be a market in the field of patent translation for what one might call high-grade crap, or slightly less inaccurate crap, which would be the logical result of applying the MT post-editing approach to patent translation.

So since there does not seem to be an easy way to monetize post-editing of machine translations in my field, I don’t think I will worry about the fact that I am leaving 100% of the MT market to Google and Microsoft.

What about your field, dear fellow-translators? Do you think that there is a market for post-editing of machine translations in your field?


Responses

  1. I wondered why I was being sent jobs from agencies who then supplied me with a “glossary” of words that had nothing to do with the subject and often contained spelling errors in both source and target language, as well as mistranslations. Then the penny dropped. These are the vocabularies they demand from translators who use CAT tools! So that is the quality of CAT!

    Liked by 1 person

  2. Yes (there is definitely a market for post-editing of machine translations in my field and language pair), … given that there are poor souls out there working for €0.05😉

    Michael

    Liked by 1 person

    • But what field and language pair is it. Polish-English (you have such a long name that I can’t be sure).

      Like

      • Ha ha, sorry: I do only Dutch to English. Various fields, mainly legal, business, education and IT, and have recently even started doing patents. Well, patent applications.

        I sometimes “post-edit” MT output, whether from Google Translate or Microsoft Translator, in memoQ via the plugin/API, or Slate Desktop, the relatively new Windows-based custom MT tool. MT can be very useful, if you know what you are doing to start with, but if anyone is going to benefit from it it’s ME, not some money-grubbing agency.

        PS: re my name: Polish grandfather, on my mother’s side (Wdowiak), Dutch father (Beijer), American mother.

        Liked by 1 person

  3. PS: FANTASTIC blog by the way, and probably the only one I read regularly, apart from Kevin Lossner’s.

    I have a little button in Chrome that sends your posts straight to my Kindle, for offline reading when I have a moment to spare.

    Liked by 1 person

  4. In English/Hebrew tech and localization world there is a market, of sorts, for post-MT editing.
    I refuse to do it.
    The amounts offered are low, and the work involved is both intense and disagreeable, and involves a different kind of work from translation. Maybe there are people out there who can create a document that conveys meaning and nuance on a foundation of approximations, but I’m not one of them. I am, however, one of the people who can look at a document that’s been MTed and post-edited and find errors by the truckload.
    Perhaps providing consultation and finding errors in translations submitted to courts as true and correct is a growth field?

    Like

  5. Thank you for your comment, Shunra.

    Similarly to Michael’s comment, I think your answer is both in the “yes” and “no” category.

    You are convinced that there is a market for post-edited machine translations and you have seen examples of it, but since it is such a distasteful job and the pay is so low, you would not want to bother with it.

    So the way I see it, the answer is “yes”, the market is there, but “no”, there is no way a human translator can make money off this market, unless it is a translator living in a place with a cost of living on par with the Black Hole of Calcutta.

    Liked by 1 person

    • Not to put too fine a point on it, it depends on the meaning of “is”.
      In other, less presidential words: there is certainly demand for such labor. And there appears to be supply of such laborers. But the end product produced is not of sufficient quality to merit the name “edited translation”.

      There are certainly varieties of work that people are willing to do without monetary pay. Such jobs mostly involve some other exchange of value: they’re fun, or fascinating, or glamorous – or necessary for survival of oneself or one’s loved ones.
      Post-editing machine garbage is not fun. It’s far from fascinating. No glamour is involved. And in terms of making enough money to survive (which is the only way it would involve survival; other work is done for survival with other, far better reasons) – it’s not a very effective way to survive.

      So yes, one could say there *is* a market, but it does not do that existing very impressively or persuasively. Over time, it would tend to cost the post-editors more money in machines and software and so forth than they would be likely to obtain from it.

      I’ll stick with old fashioned translation.

      Liked by 1 person

      • “So yes, one could say there *is* a market, but it does not do that existing very impressively or persuasively. Over time, it would tend to cost the post-editors more money in machines and software and so forth than they would be likely to obtain from it.”

        Hm, that is interesting.

        Everybody so far seems to be in agreement with you.

        Like

  6. Yes, there is for the sort of translation I mainly do (marketing, media, business, etc), and for my main language pair (EN>PT-BR). I suppose, I haven’t looked for this kind of work but I have done it although I wasn’t aware I was doing it – as I have mentioned in a post here, recently I was given a job that was clearly not humanly translated at many places so what was supposed to be editing turned out to be heavy editing because the syntax didn’t always make sense. I am still not sure if the agency simply passed on post-editing to me and didn’t tell me about it or if they were simply fooled by the translator (nobody at the agency speaks Portuguese). I guess it was the latter and I have noticed in similar assignements that some translators use MT in big jobs for agencies where people don’t speak their language because they accept pressing deadlines for high volumes – and probably charge less than I do. It was not the first time this happened so I assume that it is not only clients that use MT for cutting corners on their budget, or agencies for making a profit, translators also do which is even more dishonest in my opinion.
    I work with this agency very often and they pay OK rates agency-wise and, unfortunately, I have a child to feed and bills to pay on my own, so in quiet months I cannot afford to say ‘no’ – they pay what I charge for editing, so I cannot complain. But I don’t really like doing it, if the text is smallish, it is ok, but if it is big or huge, it is a nightmare. It really would be easier – and cheaper – if they just passed the original job for me to translate and then have it proofread. My editing also has to be proofread.
    If I am prepared to offer this kind of service? Yes, as long as they pay a reasonable rate. I edit non-translations so editing a translation is not a big deal, as long as it is something that really resembles a translation or it is fairly paid. Because in my language pair, the result of MT can be the stuff of nightmares.

    Like

  7. Thanks for your response, Daniela.

    I wonder, what is a reasonable rate for “editing” machine translations in your opinion?

    Like

  8. Hi! I guess charging a figure between what one charges for translation and what one charges for editing. Depending on the “state” of the text. I sometimes rather charge per hour , it is difficult to price work like this before starting the job because we someimes think it is going to take “X” hours and it takes “Y”. But I would charge what I would charge per translation hour because it can be as translating the text again. Like I said, last time they said to me it was proofreading/editing and that’s how I charged them, I didn’t make a fuss because they really are very consistent clients but I won’t be doing it again if it is not clearly stated the kind of work that is. Or I will simply decline it.

    Like

    • “I guess charging a figure between what one charges for translation and what one charges for editing.” Which is exactly what “the translation industry” is saying that translators should be doing.

      Like

      • I thought “the translation industry” was paying proof-reading fees (at least is what they always try to pay…). As I have said, it depends on content, time spent on project, etc – that is why I said I would rather charge per translation hour.

        Like

  9. “MT will be just as good (or almost as good) as human translation very soon (although estimates vary, from three to 20 years from now; there has not been not much change in this respect in the last two decades or so).”

    I don’t recall anyone saying in 1998 that MT will be just as good as human translation in three years, or five years or even ten years then.

    MT has dramatically improved in all language pairs. With Japanese to English, Indian translators with lower Japanese/English skills started to use Google Translate to break into translation for low rates.

    MT will just keep getting better and better. Japanese patents may be the last to fall but as other areas use MT, then presumably Japanese translators will all go to patents, thereby bringing down the average wage. J/E patent translation wages haven’t increased since 2000 and so inflation (but not if in Japan) has eroded earnings around 30%.

    Like

    • “MT has dramatically improved in all language pairs.” It has improved in the direction to English, definitely not in all language pairs. But it still produces crap. After 30 years of incremental improvement, it still produces mostly crap, even in the direction from various languages into English. In other directions, it’s unbelievable crap, as far as I and most translators whose opinion I value are concerned.

      Like

    • http://www.theregister.co.uk/2014/02/06/google_translate_issue/

      Like

  10. MT produces crap for Japanese/English, and I can’t believe there is a language pair out there that isn’t noticeably better in 2016 than in 2000. The point is that free MT has let bad translators into the market at lower rates because they get important clues from it.

    When I started J/E translation in 2002, I needed help from a Japanese friend for the toughest convoluted sentences – maybe 2% of the total. She would write back with her translation and say the problem was that the Japanese (diary entries) wasn’t good. Still, her not so good English interpretation would help. My guess is that if I had GT back then, I wouldn’t have needed her help.

    Like

    • This is absolutely true. MT has lowered the standards for entry into our profession to an extent that was unimaginable when I was starting out as a translator 30 years ago.

      Like


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: