Person-Job Matching Incorporating Preference Information From Both Parties

Online recruitment has subtly changed the traditional way of recruiting in the job market.However,the increasing volume of online recruitment data makes it increasingly difficult for recruiters and job seekers to sift through valid information,and recruitment platforms need to develop smarter and more efficient algorithms for matching job profiles.With the rapid development of deep learning techniques,researchers are applying them to the field of intelligent recruitment in order to better identify the personality traits of recruiters and job seekers.In this paper,we use an attention-based neural network model for deep mining of job texts and CV texts,and build a job matching algorithm that takes into account the preference information of job seekers and the preference information of both recruiters.The following are the research results of this paper:(1)A job-matching network that incorporates the preference information of job seekers is proposed.The current research on job-matching mainly focuses on how to improve the feature representation of job descriptions and job seekers’ CV texts,without considering the preference information of job seekers.In view of this,this paper proposes the PJFJP(Person-Job Fit Based on Jobseekers Preference)algorithm,which takes jobseekers’ job browsing records and desired job categories as their effective preference information.The algorithm model first learns a complete feature representation of job descriptions by learning at the word-sentence-text level,specifically by using the Bi GRU technique and the Word2 vec method to train vector representations of words and sentences.To highlight the key requirements of the job descriptions,the attention mechanism is used to weight the sentence representations of different job requirements to obtain the text-level feature representations.Next,Word2 vec and TF-IDF methods are applied to give textual representations of job experiences.As job browsing records can reflect the preference information of job seekers,they are stitched with the feature representation of work experience to obtain the feature representation of work experience fused with job browsing records.Again,word2 vec and TF-IDF techniques are used to obtain the desired job feature representation.Finally,the text similarity between job descriptions and CVs is performed using a multi-layer perceptron(MLP).The effectiveness of this paper’s algorithm is verified based on the smart recruitment dataset,and the results show that the algorithm improves the accuracy of job and resume matching in the smart recruitment process.(2)A person-job matching network that fuses the preference information of both recruiters is proposed.In the actual recruitment process,the potential hiring preferences of the recruiter and the preferences of the job seeker can together affect the job matching results.In view of this,this paper proposes the Person-Job Fit Based on Mutual Preference(PJFMP)algorithm by incorporating the records of the recruiter’s test candidates as preference information into the feature representation of the job description.Specifically,the algorithm learns a preference representation for a job from a historical memory module that holds information on all interview candidates,where the memory module is updated iteratively to remember job-related descriptions and add preference information to the job representation.Finally,the resume representation incorporating the preference information of the job seeker and the job representation incorporating the recruitment preferences of the recruiter are subjected to similarity calculation.The experiments show that the dualpreference job matching algorithm proposed in this paper can effectively improve the accuracy of matching jobs and CVs,thus improving the efficiency of the recruitment platform,as it takes into account the preferences of recruiters and job seekers in the recruitment process.