Crowdsourced Translations for Faster Course Creation!
I absolutely love Duolingo and have been using it for years now. I want to share an idea I have to make it even better. As I'm sure many of you are aware, the bottleneck in creating new courses is usually the course creation. In the crowdsourcing spirit of Duolingo, wouldn't it be great to take it one step further and crowdsource the translations?
From a potential contributor's point of view, it's often difficult to make a commitment of enough time to be able to substantially help a course to completion as it takes hundreds of hours to translate all the sentences. With this approach, even if a potential contributor knows he can only help an hour a week, he can still help! Because of this, there could be hundreds of contributors!
While I think it'd still be important to have a select few course contributors to establish the structure of the course and lay out each sentence, the translation of those sentences could be crowdsourced. This would remove the tedious work from the course contributors and allow them to progress quickly through course creation.
Though there isn't any official channel I could find to contact the Duolingo development team directly, I have submitted a bug report with the idea. Who knows if anything will come of it.
What do you Duolingoers think of the idea? Maybe if we generate enough interest in this approach, the Duolingo team will take notice!
I was getting ready to post this exact phrase, until I scrolled down and saw it.
Hi, check out the part where I mentioned, "I think it'd still be important to have a select few course contributors to establish the structure of the course and lay out each sentence".
I agree that too many people trying to establish a course direction and structure would make it difficult, but this approach isn't crowdsourcing that aspect-- only the tedious translation aspect.
No idea why you seem to be getting down votes, especially with that lovely attitude of yours! Although this would help build a bigger vocabulary for a language much faster, there will always be people abusing this power. Like what @Buenotc said, "Too many cooks spoil the broth". But otherwise, I think that idea of yours is pretty good. Keep 'em coming, you may be able to create some sort of serendipity in Duolingo one day! ;)
it's not a matter of "abusing power" in the traditional sense of maliciously using power to one's own ends, it's a matter of any john doe that thinks they should be translating when they simply aren't good enough...or, more importantly, it's about a course not having continuity because there are 30 contributors doing their own thing whenever they want instead of an actual team planning out the course progression.
To address your two points:
it's a matter of any john doe that thinks they should be translating when they simply aren't good enough
Look at immersion for proof that that doesn't matter. Immersion is filled with people learning a language (i.e. not even confident enough to know if their answers are correct). An independent study showed that the translations in immersion were just as good if not better than professional translations.
it's about a course not having continuity because there are 30 contributors doing their own thing whenever they want instead of an actual team planning out the course progression.
Check out the part of my idea where I mentioned, "I think it'd still be important to have a select few course contributors to establish the structure of the course and lay out each sentence". The main course contributors are alleviated from the tedious burden of translating all the sentences but not from defining how the course should work.
I highly doubt the results of that study...but since you don't reference it, there's no way for me to look into it.
There would be a huge problem to assure the quality of those translations. Duolingo courses are different from Memrise for example as certain standards are assured for each course. For example, that all accepted answers are correct and that as many correct answers as possible are accepted. This especially includes different varities of the same language (e.g. UK/US English). If a translation is accepted for a word once, you'd have to accept it in all other instances where it makes sense as well...
These kinds of standards are extremely hard to maintain if there is a big team (which makes communication way harder in general). So, all added translations would have to be checked by the contributors that a) proved their language skills and b) are familiar with the guidelines, like the error reports now. And the error reports right now are a lot of work for the course contributors, they often keep them busy for quite a time after a course was released. Allowing users to enter their often... erratic translations even earlier than beta would be a huge mess and probably even slow things down.
If you have less time than needed for being a contributor, there are other things you can do! Make interesting, informational posts about the language and the communities where it's spoken, make a Memrise course, start a language learning partnership...!
I don't think you're giving the concept of crowdsourcing enough credit for its ability to provide great translations. As I mentioned above, Immersion is filled with people learning a language (i.e. not even confident enough to know if their answers are correct). An independent study showed that the translations in immersion were just as good if not better than professional translations.
In addition, it'd be easy to post the rules to all the contributors on the contributor console (e.g. accept US and UK English, before using a new translation of a word you must ensure it's been taught before). These are the same rules course contributors establish implicitly now. I've often seen errors of the type you mention in the Spanish, French, and Italian courses, which were created with less crowdsourcing than we currently have in Duolingo.
Care to give me a link to that independent study? I never heard about something like that and my own experience with Immersion focussed on inconsistent formatting and documents never being finished. Oh, and lots of complaints about trolling.
Anyway, the goal of a course (no false positives at all with as little false negatives as possible) is very different from Immersion (the best true positive). Selecting which answers are acceptable and which not requires quite a bit more of understanding how the language works than just picking one good translation (and native speakers don't always have that knowledge! I had lengthy discussions with fellow German who insisted that their way of saying something was the only correct way until I proved them wrong with the Duden and they still didn't believe it despite the Duden being the de facto standard of German. And not just with one person but with several.)
Also, to get a translation quality similar to Immersion, it would require a similar review system which is infamous for the trolling... And things would still need to be double-checked from a selected group that proved its language skills because "probably okay translations" just aren't enough (of course, not all courses are perfect but the strategy is to have a good course when it's released into beta and then polish it further by relying on the users' suggestions). Other than in Immersion, there'd need to be a way to ensure completion of the course... And I'm pretty sure that the majority of the good translations in Immersion are done by a relatively small number of people, many of whom are active enough to be a contributor (if they knew a language that is worked on right now).
The current concept of having a small team that can communicate well and establish certain standards for the basic course layout and later on allowing users to polish things even further is really good IMO. It ensures that the basics are done tidily and and correctly in a reasonable amount of time but also uses crowd intelligence to fill out the gaps in a pretty short time. I doubt crowd-sourced translations right away would speed up significantly as there would be so much communication overhead and things would need to be re-done several times until they finally match the standards... And mind you, the courses you mention for being erratic were done before model came into action! They followed a fixed tree layout and just translated everything... The incubator courses on the other hand are done by specialised teams and can focus on a language's individual properties better which also improved course quality.