Resume Matching Framework via Ranking and Sorting Using NLP and Deep Learning

Online job search through websites has been a remarkably advantageous tool for both job seekers and employers, effectively serving their needs for numerous years. The job-resume matching system uses natural language processing (NLP) techniques to analyze the content of resumes and job descriptions. In this study, we used AI models to analyze resumes and rank each resume based on its similarity to the job description. This proposed framework uses optical character recognition and object recognition models to detect text and sections from resumes. Afterward, the text preprocessing step is applied. The preprocessed section is classified with the BERT model. The section of resumes classified into 5 different categories (education, experience, skills, personal, and language) are sent to the information extraction module. By using the named entity recognition method, the module identifies 8 different categories in the text (title, degree, major, skill, language, city, country, and date). Based on extracted sections and named entities similarity between the job description and resume is calculated using cosine similarity. Each section's similarity score is calculated independently. This allows users to assign each section a weight based on the importance for the total score. Finally, in the developed web application, these weights are loaded into the system and resumes are compared with job applications to be sorted. The proposed resume matching framework accurately extracts relevant information from resumes, displays the similarity results of the candidates by ranking and sorting, and helps recruiters or managers to select the best-scored candidates.