A HYBRID FRAMEWORK FOR RESUME-JOB MATCHING SYSTEM

Abstract

In recent years, modern recruitment processes have generated an increasing number of job applicant submissions and this is resulting in a growing need for developing a method for evaluating job applicant submissions consistently. In addition to surface text similarities used by automated matching tools and independent suitability prediction, there has been no method developed that captures how job applicants are compared during actual selection processes. This study proposes a hybrid decision-making methodology combining three components of contextualized text representations; standardized skill alignments through use of the ESCO classification system; explicit qualifications matching; and ranking based optimizations to enable the comparison of job applicants relative to one another. This study focuses on creating the model's architecture, testing the model on a data set of resumes paired with job postings and investigating how different information sources affect performance. The results demonstrate strong capability in identifying relevant candidates, stable predictive behavior and better discrimination than the baseline methodologies. Furthermore, the results also suggest that when semantic understanding is combined with structured competency constraints, a more reliable representation of hiring decisions can be produced. Therefore, this research concludes that multi-criteria ranking offers a viable basis for the development of AI-assisted recruitment systems and can facilitate the implementation of transparent and scalable candidate screening in actual organizational environments.