Quality-aware Online Task Assignment Mechanisms Using Latent Topic Model

Yang Du,Yu-E Sun,He Huang,Liusheng Huang,Hongli Xu,Xiaocan Wu

DOI: https://doi.org/10.1016/j.tcs.2019.07.033

2018-01-01

Abstract:Crowdsourcing has been proven to be a useful tool for solving the tasks hard for computers. Due to workers’ uneven qualities, it is crucial to model their reliabilities for computing effective task assignment plans and producing accurate estimations of truths. However, existing reliability models either cannot accurately estimate workers’ fine-grained reliabilities or require external information like text description. In this paper, we consider dividing tasks into clusters (i.e., topics) based on workers’ behaviors, then propose a Bayesian latent topic model for describing the topic distributions and workers’ topical-level expertise. We further present an online task assignment scheme which incorporates the latent topic model to dynamically assign each incoming worker a set of tasks with the Maximum Expected Gain (MEG). The experimental results demonstrate that our method can significantly decrease the number of task assignments and achieve higher accuracy than the state-of-the-art approaches.