Hey Skritter team

Just read that blog post about how certain people are cruising their characters using the Heisig approach. That got me thinking about how it would be great to start a small "study" on which type of studying is the most efficient. I don't know how feasible the idea is in terms of readiness of data, but I imagine that Skritter has tons of information on the progress of all the skritter users.
If it were possible to make some groups (by asking people about studying habits: do you just start writing without giving thinking that much? do you use mnemonics? Do you always the meaning of the character in a dictionary? Heisig approach?how much time do you still spend reading in Chinese? etc.) you could hopefully easily recognise some clearcut patterns and be able to provide some great hints. Although there surely is a fair amount of research for languages in general and probably also some on learning characters, this would be precisily tailored to Skritter (with it's repetition algorithms and actual writing).
I could also imagine that if you said it is too much work, you could find some student to do a paper on it (if you get a psychology student on it he or she might want to run a factor analysis or a multiple regression which would be really interesting).

Is that anything worth looking into?

Cheers

David

I think doing some mining of the Skritter data to improve Chinese learning (e.g. how to use Skritter to help learn, how to improve Skritter, or more general knowledge about how people learn) is a wonderful idea. I do hope that Skritter captures enough data (for example, the "X" button to see which characters people are overriding as being wrongly marked correct or incorrect, or squigs to see where stroke order detection could use work - 回 trips me up pretty often by recognizing the inner box while I am writing the outer box).

I can just imadgine the insights that you could come up with when you attach Tableau, SpotFire, and/or SAS to your database.

Yeah, that's a great idea. There is a ton of stuff we can do with so much data. I could spend years analyzing it. But then I wouldn't be developing!

The sweet spot for us is to identify easy-to-find truths we can pull from the data and use those to improve the site, whether it's by tuning the scheduling, recommending practice habits, or changing the interface to guide learners. We'll do those. But are limited in how fast we can get stuff done, so yeah, we'd probably have to get a student to do most of the work for us in the short term.

Which is what we did this summer with Maksym. (He did a bunch of other stuff, too, not least of which resulted in the Japanese version.) He designed a data analysis framework for improving our scheduling, which can be used to make things way more efficient.

But it's not quite done, as we still need to try a bunch of models for how reviews affect the forgetting curve before we can really start crunching the data. I haven't yet gotten enough time to think about starting working on that again. So we managed to get behind even with Maksym doing most of the work for us.

That's cool. I'll definitely be keeping my eyes open on posts from you guys ^^