The Study Of Task Recommendation Method Based On Workers' Interest And Competency For Crowdsourcing

The problem of information overload is accompanied by the rapid development of crowdsourcing platform under the open environment of Internet,which makes workers face the problem of task selection.However,task search is difficult to meet the individual needs of workers.As an effective way to solve information overload,task recommendation has been discussed and studied extensively.Therefore,based on the study of recommendation method in traditional domain and the analysis of crowdsourcing and worker behavior characteristics,this paper puts forward a method of task recommendation based on the workers'interest and competency.Supplement methods are proposed to achieve the recommendation of workers'cold start and tasks'cold start.On the basis of collaborative filtering recommendation,the first step of the task recommendation method which considering the interests and abilities of workers is building worker models.Based on the analysis of the workers'behavior characteristics,the worker model is firstly built based on workers'interest through using the TF-IDF scheme.The new worker model is built by integrating KSAO competency set on the basis of the theory of competency analysis into the above worker model.Then,on the basis of worker models,the comprehensive similarity is calculated to produce recommendations finally.The weight coefficient?_i??_i,and fusion coefficient?are introduced to calculates the comprehensive similarity integrated interest with competency by cosine similarity,Jaccard similarity and improved Cosine similarity.Then the nearest neighbors of target workers are found and the top-N task recommendation list is generated finally.However,recommendations for new workers and new tasks cannot be solved based on the above methods.Optimization of task recommendation is proposed under cold start conditions.A recommendation method based on workers'file information is put forward for workers'cold start,which combine with the registration information of new workers in crowdsourcing platform.For tasks'cold start,the idea of content based recommendation algorithm is combined,the LDA topic model is used to excavate the content theme of the old and new tasks,and then the category of the task is integrated,and the recommendation method based on the tasks'content and category is put forward.Finally,using real data from the website of ZBJ to experiment,the results showed the effectiveness of the method.The contrast experiment shows that the proposed method achieves better recommendation results compared with the traditional collaborative filtering recommendation method and the traditional cold start recommendation method.This paper enriched the task selection study of crowdsourcing from the perspectives of recommendation,and has a certain practical significance of solving information overload,enhancing personalized experience and so on for crowdsourcing.