Automatic Resume Screening with Content Matching
Merve Elmas Erdem · 2023
Candidate screening is the most costly step of the recruitment process. In order to eliminate a large number of candidates applying for a position, resumes are filtered based on features such as graduation department, technical skills, and experience. Within the scope of algorithmic recruitment, there are many studies on the automation of the recruitment process using semantic and data-based approaches. In this study, a natural language processing-based screening algorithm that automatically calculates candidates' suitability for a job posting using different categories is proposed. The proposed method generates a metric for each candidate that indicates their eligibility for the job post they applied for and ranks the candidates by using structured data. BERT based named entity recognition and stochastic gradient based domain classifier were used for the analysis of resume and job posting data. Semantic and fuzzy approaches were used together for the calculation of the similarity metric. The proposed algorithm was evaluated using manually annotated resume and job posting datasets that are filtered to include only software development categories.