Information Technology Reference
The explanation of the process:
One task attendant of crowdsourcing task A ( taa ∈ TASA ) submitted a
subtask of crowdsourcing task A ( sa ∈ SSA ) and produced a Submit(taa, sa) .
For Submit(taa, sa) form a corresponding quality testing subtask sb and put
into SSB .
One task attendant of crowdsourcing task B ( tab ∈ TASB) submitted a
subtask of crowdsourcing task B ( sb ∈ SSB ) and produced a Submit(tab, sb)
or called label l[tab][sb] .
Update label and get the new label l using Expectation Maximization
Algorithm with Separation of Bias and Errors. Then we can obtain the
correct label T(sb) for every subtasks of crowdsourcing task B and Cost(sb)
for the recoverable bias cost of every task attendants in crowdsourcing task
From the correct label T(sb) we can educe the corresponding quality or grade
of Submit(taa, sa) then get the quality evaluation of every task attendants of
crowdsourcing task A.
Generally, TASA and TASB have no intersection mainly due to the following two
considerations: (1) Task attendants only focused on one single work, will be
practiced in it soon and the cost of training is relatively low. (2) Task attendants
that concurrently involved in both tasks might be self-deceptive to judge their own
“Crowdsourcing Task Using Crowdsourcing Quality Testing Method”
Compared to “Two Crowdsourcing Tasks”
For “Crowdsourcing task using crowdsourcing quality testing method”, the first
response might be it's just “Two crowdsourcing tasks” while the subtasks of
crowdsourcing task B come from the submitted results of crowdsourcing task A. This
understanding is right if we separately look these two crowdsourcing tasks. However,
as a “Crowdsourcing task using crowdsourcing quality testing method”, it has features
and advantages that “Two crowdsourcing tasks” does not provide.
4.2.1 Task's Finish Time
The completion time of “Two crowdsourcing tasks” includes two crowdsourcing
tasks' independent working time and their transformation and generation time. While
in “Crowdsourcing task using quality testing method”, two tasks are working
concurrently. The transformation and generation time are covered in the working time
of the two crowdsourcing tasks and do not need any extra time consummation.
Therefore its completion time equals the two crowdsourcing tasks' independent
working time minus their concurrent working time.