Comparative Semantic Resume Analysis for Improving Candidate-Career Matching

Abstract

A resume, in general, is a commonly and widely used way for a person to present their competence and qualifications. It is usually written in different personalized methods in a variety of inconsistent styles in various file formats (pdf, txt, doc, etc.). The process of selecting an appropriate candidate based on whether their resume matches a list of job requirements is usually a tedious, difficult, time-consuming, and effort-consuming task. This task is deemed significant for extracting relevant information and useful attributes that are indicative of good candidates. This study aims to assist human resource departments to improve the candidate career matching process in an automated and more efficient manner based on inferring and analyzing comparative semantic resume attributes using machine learning (ML) and natural language processing (NLP) tools. The ranking support vector machine (SVM) algorithm is then used to rank these resumes by attribute using semantic data comparisons. This produces a more accurate ranking able to detect the tiny differences between candidates and give more unique scores to get an enhanced list of candidates ranked from the best to worst match for the vacancy. The experimental results and performance comparison show that the proposed comparative ranking based on semantic descriptions surpasses the standard ranking based on mere regular scores in terms of a distinction between candidates and distribution of resumes across the ranks with accuracy up to 92%.