CrowdSelect: Increasing Accuracy of Crowdsourcing Tasks through Behavior Prediction and User Selection

摘要: Crowdsourcing allows many people to complete tasks of various difficulty with minimal recruitment and administration costs. However, the lack participant accountability may entice as possible without fully engaging in them, jeopardizing quality responses. In this paper, we present a dynamic time efficient solution task assignment problem crowdsourcing platforms. Our proposed approach, CrowdSelect, offers theoretically proven algorithm assign workers cost manner, while ensuring high accuracy overall task. contrast existing works, our approach makes assumptions on probability error for workers, completely removes that such is known apriori it remains consistent over time. Through experiments real Amazon Mechanical Turk traces synthetic data, find CrowdSelect has significant gain term compared state-of-the-art algorithms, can provide 17.5\% answers' previous methods, even when there are 50\% malicious workers.