First off, I want to say how friggin’ awesome Skritter is. It has allowed me to fill in the big writing gap I’ve had with Japanese, and improved my reading quite a bit along the way. Thank you everyone involved for making the most useful piece of software I’ve ever used for Japanese learning.
I’d like to touch on the data quality on the Japanese side of the app. As an intermediate learner of Japanese, I’ve been able to identify ~10 errors with pitch accents in word recordings and character writings in the first 1500 or so words I’ve studied, maybe 30% of which don’t even have recordings. Jeremy has been extremely responsive in fixing them quickly (thank you!). But I’m sure that I’m just missing the large majority of errors, given that I’m just a learner, especially with noticing pitch accent errors and minor character errors (this line shouldn’t go through that one, etc.) – what would a native speaker or kanji expert have to say, respectively?
Ultimately, crowdsourcing data quality to a community of language learners seems like a methodology that will never ensure good data. These errors frighten me when I think about how I’m spending time learning incorrect things (especially the pitch accents – those are probably impossible to fix once you learn it wrong)!
Has Skritter considered hiring or contracting experts to clean up the data?