Research On Task Assignment For Complex Task Crowdsourcing

Crowdsourcing is a distributed solution mechanism based on Internet public intelligence and resources.In the past few years,crowdsourcing platforms such as Didi,Amazon mechanical Turk(MTurk),Airbnb,Upworks have emerged on the Internet.At present,tasks released on the mainstream crowdsourcing platforms are mainly relatively simple and independent of each other such as data annotation and image recognition.MTurk defines such tasks as human intelligence tasks(HITs).In many situations,however,crowdsourcing is required to support more complex task.This type of task contains many sub-tasks and often requires the collaboration of workers with different professional skills.In this paper,the crowdsourcing that requires multiple specialized skills and has dependency within sub-tasks is defined as complex-task crowdsourcing.When performing complex-task assignment in crowdsourcing systems,it is necessary to consider the extent of worker's applicability to the task.This paper define it as the availability of workers.On the one hand depends on the ability of the workers to complete a certain type of sub-task;on the other hand,there are certain dependencies between sub-tasks of a complex task meanwhile the assignment and the smooth completion of complex tasks need to fulfill the requirements of the dependency.This paper research on task assignment problem in complex-task crowdsourcing which has multi-skill requirements and dependency features.To assess the worker availability in complex-task crowdsourcing,an availability measurement algorithm evaluating the skill proficiency,workload,and reputation of a worker has been introduced to estimate a worker's applicability of certain type of task and increase the productivity.Based on this,Efficiency Optimal Complex Task Assignment Algorithm(EOCTA)and Assignment Number Optimal Complex Task Assignment Algorithm(ANOCTA)are proposed respectively,both satisfying multi-skill requirements and dependency features of complex task.The EOCTA algorithm iteratively calculates the availability of workers based on greedy strategy and choose optimal solution in each round.ANOCTA algorithm modeling the task assignment problem in complex-task crowdsourcing as maximum flow problem to maximize the assignment of complex task.Algorithms in this paper can effectively assess worker's applicability to the task with multi-skill requirements and dependency features,and provide practicable solutions for task assignment problem in complex-task crowdsourcing.Finally,with the background of Internet garment industry,this paper designed a crowdsourcing-based prototype platform for personalized customization and conducted simulation experiments by simulating the scenarios of the Internet garment industry.Experiment for worker availability shows the correlation between the number of tasks and the workload of workers,the number of people who undertake the tasks.Experiment for assignment algorithm compares the experimental results of the EOCTA algorithm and the ANOCTA algorithm in terms of the number of tasks assigned,algorithm running time,and efficiency gain.Experiments prove that our proposed algorithm has a good effect in solving task assignment problem in complex-task crowdsourcing.