I used MTurk for a data gathering project recently. Turnaround time is
great: within a few hours. You can set params to only accept people
whose rejection rate is 5% or less or what not (95% acceptance rate is
the default) to weed out the riffraff. Pretty cost effective.
My instructions were complex (check websites A, B, and C and on each,
look for data elements X, Y and Z and do the hokey-pokey and...) and as
a result the quality wasn't great. To avoid: I'd recommend either
farming out each subtask into a separate HIT or aggregating the entire
task in a single HIT.
--SET
Aaron Rubinstein wrote:
I'm looking at Amazon's Mechanical Turk, https://www.mturk.com/mturk/welcome
to "automate" checking the page order of 11,000 scanned books (some were
scanned backwards by a long out-of-business vendor).
Does anyone on the list have experience using Mechanical Turk? This is a
time sensitive project so I was hoping to figure out some best practices for
attracting Requestors, and maybe even a sense of how quickly tasks might get
accomplished.
Thanks,
Aaron
