How I imagine user contributed languages will work.
I have actually given it some thought myself, even before Duolingo existed, I had a general idea that there should be a way to create an automated learning system that builds step by step on previously absorbed information. It occurred to me shortly after discovering pimsleur audio courses (which are great too by the way, if you are like me and can't stand devoting time to studying grammar books...). Of course ideas are cheap, and I'm always happy when things that I have thought of magically appear out of nowhere without me having to do anything :)
So here is how I imagine how it could work:
Step 1: Create a database of the most commonly used conversational phrases and expressions in the target language, ensuring to also include phrases for all of the 2000-2500 most commonly used words. Tag every phrase according to subject, grammatical feature and level of complexity. This should be as open and flexible a database as possible, so that it can be easily improved and amended at any point, by anyone that knows the target language. This is a very difficult step, but once it is done, there is enough information to trivially build the classic Duolingo skill tree, possibly even enough that the process could be semi-automated.
Step 2: Create a system, perhaps even a different mini game, with the goal of translating each of those phrases into all the other languages duolingo supports. This creates the skeleton framework for all the learning X from Y combinations, so combinations with more demand would float naturally to the top. Obviously, phrases can be translated in a thousand different ways, something that I find constantly because I speak British English and not American English. Often translations that sound perfectly reasonable to me make me lose a point. I am sure this problem is one of the hardest ones to solve, so perhaps it will be important to meticulously monitor every incorrect answer that users give, so that the ones that occur frequently can be either accepted as valid, or given the 'tooltip hint' that explains why they are not. The experience will be very rough at first, but will eventually burn in quite nicely. One thing that REALLY needs to be added is the ability to challenge an incorrect response from your mobile device, just like the website, because without such a feature you are throwing away immense amounts of useful feedback.
Step 3: ??????
Step 4: Profit!
So am I close?
Something else also occurred to me after sleeping on it; if the system is going to involve any kind of monitoring of incorrect answers, and this might basically mean volunteer teachers keeping a watch on the back end of things, then that is a perfect opportunity to harvest not only whether a given submitted answer is right or wrong, but WHY it is wrong and how to correct it. This would be much less annoying than the usual, curt 'You used the wrong word' penalty. This 'teachers notes' commentary would only ever have to be written once per phrase, but would benefit hundreds and thousands. Of course my imaginary database is starting to look pretty convoluted now with all these ideas, but that's why I am a cook and you guys are the programmers :)
Absolutely love this idea- if duolingo allowed its users to do this maybe less 'abundant' languages like Irish, Lithuanian, Romanian and even Latin could make their way onto this site! It would certainly give the available language courses a massive boast and as duolingo only adds languages which are on top demand to be translated onto the web a whole new host of language might be made apparent http://www.duolingo.com/comment/693743