A Time-Series-Based New Behavior Trace Model for Crowd Workers That Ensures Quality Annotation

Crowdsourcing is a new mode of value creation in which organizations leverage numerous Internet users to accomplish tasks. However, because these workers have different backgrounds and intentions, crowdsourcing suffers from quality concerns. In the literature, tracing the behavior of workers is preferred over other methodologies such as consensus methods and gold standard approaches. This paper proposes two novel models based on workers' behavior for task classification. These models newly benefit from time-series features and characteristics. The first model uses multiple time-series features with a machine learning classifier. The second model converts time series into images using the recurrent characteristic and applies a convolutional neural network classifier. The proposed models surpass the current state of-the-art baselines in terms of performance. In terms of accuracy, our feature-based model achieved 83.8%, whereas our convolutional neural network model achieved 76.6%.

Automated summary

Crowdsourcing is a new mode of value creation where organizations utilize a large number of Internet users to complete tasks. However, different backgrounds and intentions of these workers lead to quality concerns in crowdsourcing. This paper introduces two novel models based on worker behavior that leverage time-series features and characteristics for task classification.