DataBounty.io is a marketplace for data processing tasks. Data scientists can submit tasks like identifying an image of a dog or cat in order to train a data model. In exchange, users are paid for executing these tasks. But what if users just enter random responses? To counter this, I introduced control questions. Each task is divided in 10 questions. Out of these, 7 are questions which answer is unknown, 3 of them are control questions which answer is known. If the user does not reply correctly to these control questions, the whole set of 10 questions is discarded and the user is not paid. If the control questions are correct, we consider the whole batch of 10 questions correct, pay the the user and proceed to the next batch of 10 questions.
Additional development will include publishing the completed dataset to the Ocean Marketplace.