Lost in Translation API

Editorial note: Here’s a guest post from Ultan O’Broin (@ultan) from the Oracle Applications User Experience Team in Dublin. You should read his blogs on translation and user experience. Enjoy.

You may not have heard, but Google is deprecating a bunch of APIs. The most prominent of these is the Google Translate API. That’s right, no more handy free translation service to integrate into your website or mobile app.

Personally, I think this is an incredible move from the Google Developer Relations folks, and the reaction from the development community was severe (see the comments). Google Translate is playing its role in machine translating lots of useful content worldwide freely and will continue, but lots of apps and sites used that API too. And, for all the joking about the quality of the translation, the result was usually pretty good enough to understand and use (provided your MO was that your life didn’t depend on it, but your vacation did), and a whole lot better than the alternative – no translation at all.

Google cited cost and abuse of the API as reason for deprecation. Perhaps it was to do with the manipulation of multilingual website SEO as hinted in the Google Webmaster guidelines. I doubt Google was running into processing problems–given their infrastructure–but who knows?

However, a bigger question remains (Google Wave fans yawn at this point) for developers: Who would trust working with these kinds of APIs again? A general problem, sure.  You could switch to the Microsoft Bing Translator equivalent, but that might go “POOF!” too.

However, I doubt we have heard the last of the Google Translate API. There is just way too much value in the data (11% of scanned printed volume?) behind it, and already some developers have pleaded with Google not to go down the deprecation path but to  let people pay for using the API. Given the massive extra effort needed to improve the quality of the major languages output even slightly now, Google needs a new approach to machine translation, one involving rules and human editors, and not just statistics. And that costs. So I expect that Google Translate API will be back, but there will be some “attractive payment options” involved if you want to use it …

Thoughts? Ever been burned by API deprecation or other, ahem, changes? How did you react? Find the comments…

PS: I love the “Don’t Shut Down Google Translate API” Facebook page. Who said geeks don’t do irony?


  1. I’ve been at places where they were unable to upgrade to newer versions of software due to deprecated features/offerings. In some demos I used some APIs from a company called Trynt which vanished very quickly. I also had some files stored on a service called LionDrive which also went pop. And links to very good Oracle articles on Dizwell.com.
    At least Google is giving some time to work out a migration path.

    There’s a risk in any external dependency. I suspect a lot of places aren’t adequately prepared for these risks. An interesting aspect will be mobile apps relying on this API. I suspect there will be a lot of people demanding refunds.

  2. I see Microsoft have put out the following statement: http://social.msdn.microsoft.com/Forums/en-US/microsofttranslator/thread/0bde21a2-b452-4dc9-94e6-1b9507b87961

    “As a side note: the Microsoft Translator API is here to stay.”

    I am sure the terms of usage can vary though…:)

  3. Given the fact that Microsoft hemorrhages money from their online unit every quarter, they don’t seem to mind keeping around products that don’t make money 🙂 

  4. Twitter is the case study now. Once the darling of developers and the beacon for how to build a development ecosystem, now the whipping boy for *gasp* wanting to control its product and make money. Developers are a fickle bunch.

  5. Interestingly, Twitter decided to use crowdsourced translations and not MT for their international versions – similar to Facebook (see Matt Sanford’s response here http://www.quora.com/What-key-benefits-did-Twitter-and-Facebook-uncover-in-crowdsourcing-their-translations). I guess, on a somewhat philosophical plane, the Human API of crowdsourcing and people ingenuity is one solution to the programmatic risks – though obviously you need to build an infrastructure for that too (in fact, Facebook are pursuing a PATENT on it… http://techcrunch.com/2009/08/26/facebook-files-for-patent-on-crowdsourced-translations/)

  6. Was that a surprise? That translation-language learning game will be sold to Google soon. Translations are a perfect fit for crowdsourcing. I wonder why Mechanical Turk isn’t used more for this type of exercise.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.