Posted by: patenttranslator | March 27, 2014

Crowdsourced Editing of Machine Translation – Another Greedy Concept Whose Time Has Come?

 

Communications between humans are more and more controlled by machines. Call an information number for just about anything and you have to listen to a machine giving you several options to choose from. Sometime without the one option that you really need – to talk to a real person.

It is much more cost-effective when businesses force humans to interact with machines instead of allowing them to talk to another human right away, and put them on hold for 5, 10, or 30 minutes so that they get hung up on while on hold.

More and more communication between humans is controlled by machines. First, it was done mostly to maximize the profits, more recently, to maximize the profits and also to spy on everybody, just in case the people who spy on everybody find something interesting about these unsuspecting humans.

********

In addition to communications between humans by telephone, machines now also control a large amount of communication that is translated. Millions of people use machine translation every day on the Internet because it is free. They may wonder why is it that the information on a website does not seem to make sense when it is “translated” by MT, but after all, it is free, so there is not much point in getting angry about it.

The latest twist is communication between people speaking different languages that is first translated by a machine, and then massaged by a crowd of anonymous humans (sometime called “engaged volunteers”) who may or may not in fact know two languages, but who definitely have nothing better to do with their time.

Sometime these engaged humans, referred to either as translators or editors, are paid a few bucks, up to 10 dollars an hour.

I would not call them editors, let alone translators. I think that a polite and respectful terms for these thousands of humans in a crowd would be underachievers.

A new startup called Unbable (thanks so much for the link, Phyllis) offers a service enabling “to integrate translation directly into the workflow, [so that] businesses can translate all of their website’s content in a flash at $0.02 per word.  …. Now translating over 30,000 words/day for over 30 customers, Unbabel’s secret sauce leverages artificial intelligence software and its stable of over 3,100 editors (or translators) to translate a website’s content from one language into its customer’s language of choice. First, its machine learning technology translates the text from source into the target language, at which point it uses its Mechanical Turk-style distribution system to assign editing tasks to the right translators, who then check the translation for errors and for stylistic inconsistencies.
Unbabel editors work remotely, via their laptops or mobile phones, on translations, which co-founder Vasco Pedro says provides the key to faster translations. This, combined with the efficiency of its task distribution and administration algorithms, provides a level of efficiency that allows editors to earn up to $10/hour.”

I am trying to visualize 3,100 human underachievers editing on their mobile phones machine translations. You probably have to be a translator to find this concept hilariously dumb. My guess is that the founder of the new service, who is now looking for seed capital in the amount of 400 thousand dollars, is not a translator. But I think that even a non-translator should know that the display on a mobile phone is so small that trying to edit a machine translation on it, while working with two languages and typing at the same time, is simply not possible. The founder of the new service probably will find the needed seed capital because people believe with religious fervor that technology absolutely can solve every problem these days, including problems with communication between people who speak different languages.

There is not a single problem in this world that cannot be solved by a clever entrepreneur with a clever algorithm.

You just need some good machine translation software, then you feed it a really large corpus of data, and then you throw the results to a really big crowd of underachievers and you will make mucho, mucho dinero in no time at all. Plus the initial 400 K, of course.

Unbable is really cool word, isn’t it? How original and imaginative! The perfect word for a service that will once for all erase the concept of the Tower of Babel from human experience in the modern world!

**********

Will the creator of this revolutionary concept be able to make money? If you remove the letter “b” after the prefix “un”, instead of the word “Unbable”, you will get the word “Unable”.

Something like that can happen very easily when you are trying to edit text on a tiny screen of a portable phone. It takes me forever to type anything on the keyboard of my iPhone because it is such a slow process and I make lots of typos (including on this blog). And that’s when I am just typing a simple sentence, without looking at text in two languages and trying to translate or edit anything. Even on a laptop, I find it hard to translate or edit. I need a good keyboard, a comfortable chair, and most importantly, of course, a good rate.

Without a really good rate, I find it basically impossible to concentrate!

It is likely that thousands of underachieving humans in a huge crowd of fellow underachievers who don’t mind working for 1 cent a word or 5 dollars an hour – the article said that the translations would be sold for 2 cents a word, and that they would be paid up to 10 dollars an hour, which probably means that they would be paid 1 cent a word or that the real rate for them would be 5 dollars an hour – will make a lot of typos like this. Lots and lots of words like “Unable” when what they really meant was “Unbable”.

When customers who want to buy something from a website are fed machine-translated text that is then edited into one huge typo on portable phones by a crowd of human underachievers, they just might get mad at being treated this way and go to a site that they can actually understand, which would be really bad for business.

I have a feeling that after this bold adventure in MT/crowdsourcing burns through the first 400 K, Unbable will need more money again pretty quickly.

Here is a thought: controlling communication between humans by machines for higher and higher profits may have reached its limit. People are getting fed up with the whole concept. I know that I am. Could it be that there is a lot of money in a new and revolutionary approach to communication between humans: instead of relying on the machine-human interface, why not use the human-human interface as much as possible?

If you have cleverly written text that is designed to make people buy a product or a service, people who have the choice to buy it or go somewhere else, have this text translated by a real translator who specializes in that sort of thing.

It may cost more money initially, but as the saying goes, you have to spend money to make money.

 

 


Responses

  1. Where is this world going? It’s not just the typing on an smartphone keyboard that is a problem. The second you hit that Enter key, you find that unbeknownst to you it has sent total crap into ether. But then the ultimate question that many translators like us are faced with, again and again, is: does translation quality still count?

    Like

    • *keypad I guess would be the proper word.

      Like

  2. @Nadine

    To hell in a handbasket.

    Like

  3. I love the Unbable concept, Steve. It is the best thing that can happen to the translation “industry”. Imagine–an MT translation, itself a heap of garbage, is thrown at a crowd-sourced “translator” or “translators” or “editors” whose capabilities no-one has checked or tested. They produce more garbage because they want to make a fast buck and nobody is going to QA their output. At a measly ten or five dollars an hour it is obvious they aren’t going to bend over backwards to assure quality. The result must be absolutely hilarious. And once the sh..t hits the fan (i.e. the client becomes aware of the garbage he has bought), he will NEVER AGAIN touch Unbable, or MT for that matter, with a ten-foot pole. So I think we should wish for Unbable to get their seed capital rounded up fast; the faster, the better. At least they will disappear in the abyss of translation history that much sooner. And yet another dream of automated translation will meet its demise.

    Like

  4. Ah, thank you for cheering up my day with your inimitable humorous take on the idiocy that reigns in the translation business.

    Like

  5. @Volkmar

    Amen.

    @Elisabeth

    My pleasure and privilege.

    Like

  6. There’s an American agency I signed up with by mistake that does crowdsourced translations. You sign in to the site and their interface has broken the text up into individual sentences, and then the translators who are logged in have to compete to grab the sentences and translate them as fast as they can before someone else gets them. I thought this was about the worst way possible to do translations, but this system you describe does look even worse.

    Like

  7. @Christopher

    And how much are they paying for something like that, I wonder.

    Like

  8. Can’t believe anyone can use a smartphone in their translation work. They must be kidding!

    Like

  9. @Unbelieving Thomas:

    That’s what the article says. You can click on the link in my post.

    Like

  10. Crowdsourced Editing of Machine Translation sounds like an incredible perversion…laymen trying to make a masterpiece out of number of randomly connected syllables.

    Like

  11. @ Alia

    Now, that’s unfair to MT.

    MT does not result in randomly connected syllables.

    Only randomly generated and randomly connected words.

    Like

  12. Oh, yes. But that popularity of crowdsoursing makes me scared. What about a crowdsourced surgery operation next?

    Like

  13. @Alia

    “What about a crowdsourced surgery operation next?”

    That is unlikely to happen. People are not that stupid when it comes to surgery, I think.

    Last week I translated an operation manual for a micro-surgery device from Japanese to English.

    Crowdsourced translation of an operation manual for a micro-surgery device, that may happen at some point. People are that stupid when it comes to translation, I think.

    Like

  14. Superb and so wittily presented!. . .Again! Thank you for continuing to serve as a beacon of truth and justice in our beleaguered industry!

    Like

  15. I find this stuff so alienating that,for the first time ever, I was only able to skim through your blog today. 😦

    Like

  16. Well, the owner of a certain German agency specializing in medical translation had this to say about the service:

    “Interesting results. There are a few serious mistranslations, but to be honest, if I could get a MT system that can deliver this quality, I would be prepared to use it. I would use it as input for my own translation – kind of post editing. There are sentences that are ok, some only require minor adaptation. It might be a tool to improve my productivity. They do offer bulk discounts. If this means that I could get my translations pretranslated in this quality for let’s say 0.015/cent per word, why not.”

    I put that in context by seeing what he considers to be acceptable language for his own web site 🙂 I’m sure of course that this professional service takes care to maintain full confidentiality for all texts submitted, so there will be no difficulties with confidential medical records for example….

    Like

  17. I think the problem is that a lot of people certainly recognise that they have a real translation need but they see the ‘language barrier’ as something that mostly gets in the way of what they’re really passionate about (their business, their activity, etc.). So they are naturally not very passionate about translation and aren’t happy to have to spend money on it. They do not see it as something valuable that supports them and it is not core to what they do. It is still a mystery to me why so many people see commercial translation as something very straightforward. Apfel = Apple. Somebody once told me (real-life encounter, a language teacher even, although I think it was more of a sideline job for her) that I’m not a literary translator, so anyone can do my job.

    Like

  18. @Kevin

    Hello Kevin, I find your story interesting because I personally could never work that way (i.e. modify and improve a bad pre-translation). If the pre-translation has “serious mistranslations”, the style and flow must be awkward too. it would get in the way of my thought process and prevent me from getting a wholistic feeling for the whole text and the flow of one part to the next. At least, that was my impression once when I had to fix someone else’s translation once that was returned by the client. And that was not done by a crowd-sourced machine. I just re-translated the thing from scratch because I found it simpler that way. I do a lot of legal and marketing though, so maybe it’s not so bad when you have something technical/medical.

    Like

  19. Phyllis, it’s every bit as bad for technical/medical. The fact that many of the people (including translators) in those domains can barely construct a sentence in any language doesn’t change that. They just figure that a software program which can approach their own level of incoherence and inaccuracy is good enough for the rest of the world.

    Like

  20. And maybe it actually is good enough a lot of the time for most people. I have encountered the mindset of, we are German, or Russian, or whatever, so of course we don’t speak or write English properly, everyone will understand that, we don’t need translation. It just depends on your needs and how you see them I suppose.

    Like

  21. Well, I posted a long reply to that article, explaining how companies would be taking a big risk using that service and my comment was deleted. That speaks volumes about the validity of the other comments and the very purpose of the article itself.

    Like

  22. @Jeff

    You mean the article posted at:

    Unbabel Launches A Human-Edited Machine Translation Service To Help Businesses Go Global, Localize Customer Support

    ????

    So they censor comments that they don’t like at techcrunch.com?

    Now, if they can figure out how to also erase blogs, their system will work perfectly, just like the newspapers Pravda and Izvestiya used to work in the former Soviet Union, until it collapsed, of course.

    Like

  23. If we all refuse to post-edit, that is, take care of those “minor adaptations” (!!!), with all the obvious and subtle consequences that has, MpT collapses also. Thanks for another great posting, Steve,

    Like

  24. Thank you, Aurora, for commenting.

    Like

  25. Steve, check out this new “start-up” https://tran.sl/
    They specialize in “human powered translation” and post jobs on an open dashboard for all to see. Translators can press on the ACCEPT JOB button at any time, but the longer the job remains available, the price, or as they call it, the “reward” goes up (to a maximum limit). It’s like the game of “chicken” for translators.Translations are edited and evaluated by a council of elders. Poor translations get yellow carded and you lose -500 reputation points, a second mistake earns you a red card with the loss of -1500 reputation points and possible expulsion. You can exchange your reward points for real cash via paypal or bitcoin and if you rack up enough reputation points, you can join the council and judge other people’s translations.

    Like

  26. @ Jeff

    “No-bullshit, badass i18n

    We built tran.sl because we could not find any decent translation solution for our software. We just wanted to have our software sound natural in many culturas”

    It will work in many culturas, I’m sure.

    Does the .Sl abbreviation mean Sierra Leone, Sri Lanka, or Slovenia?

    I should work at least in one of these culturas.

    Like

    • Shocking.

      This is from their own website:

      “Please do not use zzzzz to translate legal contracts, patient notes, patents, technical documentation, manuals or academic engineering documents. Those documents have an extremely slim margin of error and some of them can only be translated by sworn translators. Those services are expensive and take a long time, but it is the only option to get the high quality, consistent translation those texts require. We encourage you to find individual professional translators in zzzzz for your language pair using our reputation ranking system.”

      Reputation ranking system??

      Among the list of examples best-suited for their system, one reads:

      > “I have a presentation tomorrow in English and I need to translate my powerpoint and a script”

      Well then, good luck with it! 🙂

      Like

  27. I just found this list of translation start-ups:
    https://angel.co/translation

    Like

  28. I thought it couldn’t get worse but it seems it can. Happened to come across this today, thought you might be interested 🙂 http://bit.ly/1mKTwSq

    “Google translation is the most well-known translation service, but it has been unable to shake inaccuracy issues. Nobody uses Google for translating business documents. We believe we can outdo Google in this field.

    Flitto is a great way to make money with your free time. While you’re sitting in the toilet, you can utilize the time efficiently by translating and earning points. With the points, you can later buy items from our e-commerce store.”

    Like

    • Thankfully they’ll get flushed at some point.

      Like

  29. @Phyllis

    It is much easier to shake inaccuracy issues while you are sitting in a toilet.

    Everybody knows that.

    Like

  30. To trace the source of a trend, one should take the advice of Deep Throat (at least as depicted in the film All The President’s Men) and follow the money.

    It seems that at least some investors have started to lose faith in the MT-cult and are now migrating to the its parent cult, the technology-cult, while funding a new contrived and convoluted business model that they call “automation of human translation”, that you so accurately described, Steve, by coining the term Nanolation.
    After trashing the reputation of MT as a tool (mainly for discovery of content that really needs translation and a professional aid in some scenarios) by trying to portray it as a human replacement, which is of course not what it is designed for, the technology-cult now seems to be determined to do just the same thing, while learning from past marketing mistakes and now highlighing the alleged “human” element involved in the the decision making part, while all along being managed by their new brilliant, advanced, sophisticated, and magical load-balancing like algorithm that takes care of everything else. A combination of a human and machine working together, just what the world at this point in time (which I like to describe as the middle-ages equivalent of the digital era) is so infatuated with.

    Charlatans have always existed and will probably always exist.

    Like

  31. “A combination of a human and machine working together, just what the world at this point in time (which I like to describe as the middle-ages equivalent of the digital era) is so infatuated with.”

    This human-machine organism working as one, kind of like three musketeers, is not really a new idea.

    Ray Kurzweil has been writing about it eloquently for about three decades now, although the technology envisaged by him has not been realized so far.

    I think that a problem with these inventors of new MT-human systems is that they know only one language and have no idea what is involved in translation.

    Incidentally, would you like to write a guest post for my silly blog?

    I really enjoy your comments.

    Like

  32. I will be honored.
    Thank you very much for the offer.

    Like

  33. […] Innovative new startups, and new ones are being created as I am writing these words on my silly blog, are based on the concept of crowdsourced editing of machine translation, namely the idea that the minor imperfections of machine translation, such as when they make no sense whatever, can be corrected by humans who don’t really need to be translators at all as long as they have some knowledge of two languages, and who should be able to make these corrections on their cell phones for something like 1 or 2 cents per word. […]

    Like

  34. […] who can drive although they are not taxi drivers? There are attempts to achieve exactly that. As I wrote in a post 3 months ago, the new efforts to eliminate human translators are based on the concept of crowdsourced editing of […]

    Like

  35. […] some knowledge of a second language can be a translator, and translating can be done easily even by a dude sitting on his throne in the bathroom, pecking away on his cell phone. People like that would deliver even higher profits than the […]

    Like

  36. […] There are crowds of people in this world calling themselves translators who are hungry for work and …. But the problem is, since only a very small percentage of them can do the job well, it’s very hard to replace a good translator when a complicated project suddenly becomes a very urgent task and the translator who used to do this project very satisfactorily suddenly becomes unavailable. The relative irreplaceability of translators then becomes very evident for example when a translator dares to take a few days off and go on vacation, although it is something that, unlike translators, most other people can do every year without worrying much about anything. […]

    Like


Leave a comment

Categories