众包模式下基于参与者胜任度和接受度的任务推送模型

A Model for Task Recommendation in Crowdsourcing Based on Participants' Competency and Acceptance Degree

任务推送是众包模式下任务执行的重要方式,对提升任务完成质量以及提高参与者和发包方满意度具有积极影响。然而,当前研究将参与者胜任度作为确定合适参与者的主要指标,较少考虑参与者接受度对任务推送的影响。基于此,建立综合考虑参与者胜任度和接受度的任务推送模型,在模型中提出识别潜在参与者的思路和过程,利用启发式相似度算法测量任务相似度,描述并量化参与者表现,进而提出参与者胜任度测量方法。同时,基于参与者以往参与任务的特点,利用扩展的粗数方法,确定参与者接受度,并综合参与者胜任度确定潜在参与者的优先序。以一品威客平台为例,说明任务推送模型和方法的应用过程和有效性。研究结果表明,提出的模型和方法能够有效弥补常用的自选择任务执行方式的不足,避免参与者从大规模任务列表中选择合适任务,缩短任务的平均完成时间,同时能保障一定的参与者数量,为发包方提供多个较高质量的备选方案。与仅考虑参与者胜任度的任务推送相比,可以将任务推送给能够胜任且愿意接受任务的参与者,在一定程度上能够保证参与者的回复率,提高任务推送的成功率,减少多次推送以及由此带来的成本和时间的增加。研究结果丰富了众包模式下参与者胜任度和接受度的测量方法,扩展了众包任务推送领域的研究成果,为后续解决多任务指派优化问题奠定基础。实践中,参与者应当多参与任务且具有较好的表现,才能进一步提高后续获得合适推送任务的可能性。同时,对一些重要且要求较高的任务,发包方可采用任务推送方式完成该任务。

In crowdsourcing systems,task recommendation is an important way to perform tasks.It has positive impacts on the quality of task completion and participants’and requesters’satisfaction degree.However,current studies have mainly concentrated on whether participants complete the tasks needs to be recommended or not,but neglected whether they accept the recommended tasks.To overcome the weakness mentioned above,this paper constructs a model for task recommendation in crowdsourcing considering participants’competency and acceptance degree for a new task.In the model,we firstly proposed an idea to identify the alternative participants for the task needs to be recommended.By using the heuristic similarity measure approach,the similarity between the new task and its similar tasks are computed.Meanwhile,this paper describes and measures quantitatively participants’behaviors and performance on the tasks that they have contributed to.Then,the similarity between the recommended task and its similar tasks and participants’performance on these similar tasks are aggregated to determine their competency.Additionally,on the basis of tasks characteristics that a participant have involved in,this paper proposes an approach to determine his acceptance degree for the recommended task using the expanded rough number method.Finally,to demonstrate the implementation process and performance of the proposed approach,an illustrative example is conducted on epwk.com,a widely used Chinese online crowdsourcing market.The experimental results show that the proposed model and approach in this paper can fill the gap brought by the self-selection principle,which is a common way of performing tasks posted in crowdsourcing.It is also helpful to shorten the average time of completing the tasks,and to avoid selecting the proper tasks from a large number of alternative tasks for participants.Moreover,the model contributes to assure the number of participants who would participate in the recommended task and provide a few alternative solutions with high quality for requesters.In addition,comparing with task recommendation considering participants’competency solely,the proposed model and approach would be better in terms of participants’response rate.Furthermore,it promotes the success rate of task recommendation and decreases the probability of repeated recommendation that would lead to cost much more money and time.The study enriches approaches of determining participants’competency and acceptance degree for a task needs to be recommended,and expands research outcomes in the domain of task recommendation in crowdsourcing.Additionally,the research results in this paper establish the foundation for further solving the optimization problem of multi-task assignment.On the other hand,our study has implications to both participants and requesters.Hence,participants contribute to much more tasks and have high performance,then the possibility that they receive the proper recommeded tasks would be larger.For requesters,they would adopt the principle of task recommendation rather than self-selection principle to perform their posted tasks,especially for the tasks with great importance and requirement.