Design And Research Of Hybrid Position Recommendation System Based On Knowledge Graph

In recent years,with the rapid development of information technology,enterprises from all walks of life have gradually started to operate online.The development of online services has caused a sharp increase in network data,which in turn has caused the problem of information overload.In the field of job hunting,job seekers often need to spend a lot of time and energy on the recruitment websites rich in massive job information to find the job positions that meet their requirements and expectations.In addition,in terms of job recommendation,new job seekers have the problem of no user behavior and relatively sparse job rating information.These problems will lead to a decrease in the accuracy and recall of job recommendation.Therefore,for job seekers to recommend jobs online,a hybrid job recommendation system based on knowledge graphs came into being.The hybrid job recommendation algorithm in the recommendation system calculates the degree of relevance between the job seeker and the job through the relationship between the resume of the job seeker and the position and the user behavior of the job seeker,and then puts the relatively high degree of relevance to the job seeker Jobs are recommended to users.The main tasks of the recommendation system are as follows:(1)This paper proposes a method for identifying entities based on natural language processing,using the BERT pre-training model and the Bi LSTM-CRF model to extract knowledge about job requirements.First,use the BERT model to encode the original job-seeking requirement information to obtain semantic representation;then,use the Bi LSTM-CRF model to mark the encoded job-seeking requirement information to identify important entities such as ability,personality,and skills.The identified entities can enrich the job domain knowledge base of this recommendation system.The data set used in this article comes from the public information of the recruitment website,and this article manually labels the crawled job requirement information.Under the manually labeled data set,this paper compares multiple entity recognition models.Experimental results show that the model achieves higher accuracy in entity recognition than the rest.(2)This paper proposes a hybrid job recommendation method,which not only considers the correlation information among job seekers’ resumes,positions,and job seeker behaviors,but also considers the impact of time factors on job rating data.This recommendation method uses Trans E to obtain a job similarity matrix that contains user resumes and job semantics.Secondly,the recommendation method uses a time decay function to dynamically reduce the weight of job ratings,and uses a latent semantic model to solve the data sparsity problem of job rating matrix,and then calculates a job similarity matrix that includes user behavior and job semantics.Finally,this paper fuses the two semantic matrices to predict the unrated positions of job seekers,and obtains the Top-N recommended positions for each job seeker.According to the experimental results,the hybrid job recommendation method based on knowledge graph proposed in this paper is 2 percentage points higher in accuracy than the contentbased job recommendation method.