Research And Implementation Of A Crowdsourcing System Supporting Task Routing

Crowdsourcing represents the act of a company or institution taking a function once performed by employees and outsourcing it to an undefined network of people in the form of an open call. With the widespread dissemination of Crowdsourcing, many Crowdsourcing systems appear on the Internet. These systems support Crowdsourcing activities by connecting requesters and workers online, thus improving the transaction efficiency of knowledge and labor.Current Crowdsourcing systems usually adopt a pull methodology for task assignment based on personal intention. This approach does not guarantee that the task is performed by the most suitable workers, resulting in low-quality results. To address this problem, we propose a different task assignment approach, which is mainly based on push methodology and partially based on pull methodology. Our approach carefully selects which workers should perform a given task, while workers have the right to refuse a task assignment and pick their preferred tasks. Meanwhile, we design and implement a Crowdsourcing system supporting task routing. Our Crowdsourcing system can fully support Crowdsourcing activities with user management, task management, contribution management and workflow management functions. User management functions include user registration, user login and profile update, enabling the system to manage Crowdsourcing participants efficiently. Task management functions include task generation and task assignment, enabling the system to distribute Crowdsourcing tasks reasonably. Contribution management functions include contribution collection, contribution selection and bonus payment, enabling the system to protect the interests of Crowdsourcing participants effectively. Workflow management functions include workflow monitoring and workflow adjustment, enabling the system to fully control the execution of Crowdsourcing tasks. In our Crowdsourcing system, we use ARS (Active RankSLDA) model based task routing algorithm to push Crowdsourcing tasks.ARS model we present in this paper extends the RankSLDA (Rank Supervised Latent Dirichlet Allocation) model by considering user activity level. This model can rank users according to their expertise and activity levels for a given task. ARS model not only learns users’ expertise from users’ scores and tasks’ topic distributions, but also extracts features from users’ activity sequences as users’ activity levels. The experimental results show that ARS model based task routing algorithm outperforms RankSLDA model based task routing algorithm.