1. Forum
  2. >
  3. Topic: Duolingo
  4. >
  5. Immersion articles need bette…


Immersion articles need better vetting

In the last few days I've encountered several articles for which there are already high quality English translations. Worse yet, some are originally written in English and translated into Spanish before various Duolingo users have submitted the Spanish versions to DL. Here's the latest example that I've encountered:

Perhaps DL should consider implementing some sort of vetting process for user submitted articles. Perhaps submissions could go into a queue, and users could have the opportunity to vote on their appropriateness (content, pre-existing English versions, etc.). A user submitted article wouldn't be available for translation work until it had cleared some reasonable number of positive votes. To encourage participation, it might be necessary to reward users with a coin or a Lingot or some other incentive. That's just a thought off the top of my head, of course. I've no idea if that's workable, but something really needs to be done to eliminate the clutter in the immersion section.

Do we really need to re-translate Aesop's fables (another offender I've seen recently).

October 3, 2013



Yes, and that's not the only time it's come up. I've posted in discussions on the subject before, and I and others will probably post about it again in the future. It's an issue that needs DL's attention, and has not yet received any. The squeaky wheel, you know.


There are many, many squeaky wheels :). But the "more languages!" wheel seems to squeak the loudest, which is probably why they're greasing that one first.


Indeed. Plus this is a problem that will eventually sort itself out, I think, as DL grows its base of paying customers (though that may take quite a while yet). As translation clients begin to push content into the translation system, random user-submitted stuff will be pushed aside to make way for revenue generating content. Meanwhile, it's still frustrating to find so many poorly chosen articles.


Reasonably enough, Duo isn't sharing details of its business with us, such as how much longer it can run on venture capital and when it needs to show a profit. It seems to me that in order to survive in the long run they need to demonstrate at some point that they can bring in a sufficient number of paying clients and that they have skillful and willing translators and a workable system under which the translators can operate. Apparently the teaching aspect is getting the attention now in order the build up the pool of translators. That is a little frustrating for those who think they are ready to do translations now but can't get interested in doing Nokia ads. Presumably the translating side will get more attention later, but we don't know when or how. The Duobot seems to be a step in that direction, though I'm personally not optimistic about it.


Some details were recently shared in this Forbes article:


Some points from the article:

  • duolingo has 5 million monthly active users
  • duolingo currently costs $500,000 a month to run
  • and tantalisingly, "Duolingo has two paying clients – large, American media, companies whose names Von Ahn cannot disclose for another three weeks."


If translation is the business model, I'm surprised the Duo team hasn't put more effort into the translation interface. Right now its pretty crude, there's not a good mechanism for upvoting or downvoting translations and there's no good way to discuss translations either.


I think we need to be patient, its only been a year or so. But there are two main problems in my opinion with respect to immersion:

  1. Lack of interest in translating of the majority of those 10 million users.
  2. The closed nature of duolingo, not having an API, limits the innovative ways to translate content.

I think part of the solution is importing lots of articles en masse(e.g. from Google Books, Wikipedia, and so on) to ensure a better pool of more interesting articles. The trick is to make translation fun, or seamless, and not like a chore.


The sore point for me is how good crowd sourced translations are. It seems that Duo is betting a lot on this down the road, and perhaps they have evidence that I haven't seen. I'm convinced that if the material is at all technical it can't be translated by people who wouldn't understand it if it were written in English, regardless of how many there are.

Learn a language in just 5 minutes a day. For free.