Research For Understanding Consumption Intention In Crowdsourcing Services Based On Representation Learning

With the development of internet technology,the popularity of crowdsourcing services has continued to rise,and a large number of internet crowdsourcing service platforms have emerged,which greatly facilitate people's work and life.However,given the variety of services and content,how to allow consumers to quickly find services that meet their expectations become a challenge for crowdsourcing service platforms.In the context of increasing information resources,search and recommendation technology has become the main way for people to obtain information,it can quickly locate and return effective information according to user needs.As a key technology of next-generation search engine,knowledge graph has played an important role in the field of search and recommendation,it can semantically extend the user's search keywords to return richer,more comprehensive information.At the same time,representation learning is an efficient method applied on the knowledge map.By mapping the knowledge base to a lower dimensional vector space and performing knowledge representation learning,we can achieve quick search on scale knowledge maps.The purpose of this article is to combine representation learning with knowledge graph to quickly understand user's intention and return expected services in the crowdsourcing service domain.The main research content of this article is as follows:(1)Realization of the represents learning model and the validation of the results.Using Fiverr website as data source to build a crowdsourcing service market,and then build a knowledge base in the field of crowdsourcing service based on existing data.According to our study on representation learning,several models(Trans E,Trans H,Trans R),which are more typical and widely used,are implemented and applied in the knowledge base,and finally design experiments to compare them.(2)The transformation of user needs and service recommendation.To recommend services based on the results of the previous model implementation.Firstly,the expression of user needs should be transformed into a form that can be applied to the representation learning model,then the characteristics of the knowledge base in the domain are analyzed,and the existing knowledge base is complemented based on the representation learning model,and some common ambiguity is expressed and the ambiguity treatment plan is given.Finally,it returns the expected service in the model according to the transformation form of user needs.(3)For the incremental data training program and implementation,Given the definition of incremental scenes,in order to make the model adaptable to the continuous updating of data under the actual scenario,the paper proposes a training program for incremental data and improves it on the Trans H model and implements the solution.Finally,the same experiment was designed to verify the effect of the improved model.Ultimately,we need to combine the results of the above steps into a complete service recommendation system that automatically generates recommendations based on fuzzy requirements.