Research On Computing The Fitness Of Person-Post Based On Ability Evaluation And Semantic Similarity

As we step into the age of Internet,the industry of information has experienced a rapid development,making enterprises recruiting people through Internet and the job seekers submitting their resumes by the means of electronic documents.Therefore,There are many Online recruitment platforms which accumulated a large amount of resumes and job descriptions.As we know,recommendation system is a big part of online recruitment platform,and when it comes to recommend resumes to the post,the content-based recommendation is a basic method to achieve it,and has a great significance.When we are talking about recommending resumes to the post,the first step is to classify the job descriptions and resumes correctly,then make a recommendation.Consequently,classification and recommendation are the crucial part when computing the fitness of person and post.Our paper takes the resumes and job descriptions as the research objects,and carries out research on the problems of high accuracy of classification and the unsatisfactory performance of content-based recommendation during the online recruitment process.The main tasks are as follows:(1)Extracted the domain dictionary and applying it to classify resumes and posts using the methods of traditional machine learning and deep learning respectively,and achieved a higher accuracy,at the same time,concluded that NB achieves the most fabulous performance by constrasting their actual results.(2)Proposed a method which based on ability evaluation to compute the similarity between person and post since the attributes of resume is scalable and measurable(called AES).AES is based on the competency model,and refines the resume's attributes as well as establishing the standards of grading and weight coefficience,then make a mark,the higher the score,the more matching of person and post.(3)Proposed a method which based on semantic similarity to compute the matching between person and post because of the semantic relevance of resume and job descriptions(called SLSS).SLSS is based on word embedding,it computes the similaritiesof kinds of word's category,and takes a comprehensive consideration to measure the the similarity of person and post.(4)Proposed two methods of hybrid recommendation which combines the AES and SLSS on the basis of AES and SLSS,one is AES-AESR and the other is SLSS-AESR.During the experiment,it was found that the recommendation based on AES and SLSS alone was not quite ideal,Therefore,a attempted strategy which based on cascade to mix the two method is adopted to boost the recommendation performance.Experiments show that SLSS-AES achieves the best recommendation performance since it firstly filtered the unrelated resumes based on semantic relationship,then recommend resumes to post based on ability evaluation.