Research On Person-post Matching Model Based On Knowledge Graph And Bert

At present,both job seekers and recruiters are still faced with the problem of "adverse selection" caused by asymmetric information,and the cost of enterprise recruitment is still relatively high.During the epidemic period,the accurate person post matching function is particularly important for online recruitment platform.In the past,the research on person post matching mainly focused on the judgment of human resource experts on keywords and the construction of manual features.However,it is concluded that person post matching methods often can not make full use of the knowledge and skills relationship described by job and resume,and can not express the semantic information of the text,which will affect the effect of job recommendation on recruitment platform,and also affect the Job-Person recommended by job seekers The trust of the job seeker even affects the job seeker's recruitment choice.Therefore,it is urgent to solve the problem of accurate model matching.In this paper,combined with the data generated by the actual recruitment,according to the downstream task demand and the actual recruitment business analysis,we build the knowledge graph of supply and demand,simulate the actual recruitment process,and build a Person-Job Fit model based on attention mechanism.Among them,the method of knowledge representation is adopted,and the knowledge graph for job hunting is represented by low dimensional embedded space,which makes the text semantic information combined with graph knowledge as the input of the downstream person post matching model to realize the full learning of the model,so as to realize the matching degree prediction of person and post information,and proves the excellence of the model design concept in a large number of experiments.The main research and contributions of this paper are as follows1.In order to integrate and mine the relationship information between the two parties,this paper realizes how to build a knowledge graph for job hunting based on e-resume and text data of e-recruitment requirements,which includes data processing,conceptual layer design of the map,information extraction and knowledge matching.In addition,knowledge extraction also introduces how to label the data and select the model.2.The existing e-resume and text data of e-recruitment requirements are stored in Neo4 j graph database and visualized.Among them,including statistical analysis of resume and recruitment requirements data,and knowledge retrieval from visual knowledge graph operation map database.3.In order to simulate the attention process of actual recruitment,a person post matching model based on pre-training BERT model and attention mechanism is proposed.Based on the method of BERT coding,the resume and job text information are encoded,and then combined with the embedded knowledge vector,they are input into the person-job fit model,so that the model can be built in the output downstream of BERT.Among them,this paper combines the recommender system method,proposes an algorithm based on attention mechanism to enhance the attention features between the job and resume,and integrates the historical optimal resume matching,so that the model can fully learn the enhancement information between the candidate's characteristics and historical characteristics,so as to enhance the performance of the model.The experiment also shows that compared with the existing person post matching method,this method has more outstanding evaluation effect,and provides a new idea for the research method of person post matching.