3 ms·
we scraped a huge amount of data, then hired people to clean it up. it does scale better down in south america where wages are quite a bit cheaper but it doesn'
by rubyrescue 17y ago
we scraped a huge amount of data, then hired people to clean it up. it does scale better down in south america where wages are quite a bit cheaper but it doesn't scale enough. i would say the difference is i can hire someone here to do some data gathering and quality control on-site whereas i'd have to outsource to mturk and get lower quality if we were based in the states.
however, since we're converging on a model that's slightly further toward the 'curated' side of the spectrum than a loopt/gowalla/nextstop, i think we're OK with doing some manual editing of things like open hours, and leaving the rest to the community - all of our community editing pieces aren't out yet (we all still have full-time jobs and/or ambitious clients)...
I love the check-in statistical analysis concept for data generation or at least quality control. very cool for higher volume locations.