In my continuing effort to leverage Mechanical Turk to help with accessibility tasks, I've done another experiment on Mechanical Turk. The full results of the experiment are posted on MIT's Deneme blog.
The main results are that turkers seem to be around 24 hours a day and are relatively quick - took less than 100 seconds on average to get an answer back. Now, we'll have to see if we can play some tricks to make this even faster.
In related news, I heard a rumor from a reliable source that WebVisum might be responsible for at least some of those CAPTCHAs that you see posted on Mechanical Turk. This is somewhat surprising to me, since I would have expected them to use a free volunteer-based model. In any case, it's still not clear to me how it works.
The main results are that turkers seem to be around 24 hours a day and are relatively quick - took less than 100 seconds on average to get an answer back. Now, we'll have to see if we can play some tricks to make this even faster.
In related news, I heard a rumor from a reliable source that WebVisum might be responsible for at least some of those CAPTCHAs that you see posted on Mechanical Turk. This is somewhat surprising to me, since I would have expected them to use a free volunteer-based model. In any case, it's still not clear to me how it works.
