Information Technology Reference
In-Depth Information
to automatically complete. At the same time, every word or phrase's translation task
is a subtask of compiling the cloud computing dictionary and one task attendant can
only finish part of the compiling work. Thus, this is a typical “collaboration
crowdsourcing task” and its suitable to apply the “large-scale automatic non-computer
testing method”.
Fig. 1.
Overview Structure of PDCCP
As showed in the Fig.1, the basic crowdsourcing task is translation. Translators as
task attendants get words from database and submit their translation results online.
The interesting part of PDCCP is that we used another crowdsourcing task for the
quality testing method part of translation results. The auditors with the translation
results as their subtasks submit their judgments as audition results online. Then we
applied Expectation Maximization Algorithm with Separation of Bias and Errors to
control the quality of auditions. The final audition results serve as a feedback of
translators work quality. If the words pass it will be in the professional dictionary,
otherwise it will give back to the translators to redo the translation.
The relationship of these two crowdsourcing tasks is that the auditor's
crowdsourcing task works as the non-computing automatic testing method for quality
control of translators' crowdsourcing task. This also provides a dynamic working
mode that can apply different task distribution strategies to stimulate task efficiency
and quality.
4
Quality Testing Method in PDCCP
For creative crowdsourcing tasks, computer might not be able to complete task quality
testing or the cost will be too high. In this situation, human testing must be applied.
However, due to the large-scale of most crowdsourcing tasks, the cost of human
