Semantic Interpretation in Job Recommendations: Ontology-Driven Resume Parsing for Personalized Job Recommendation Systems

Abstract

In this study, an ontology-based Hybrid Job Recommendation System (HJRS) has been proposed to facilitate job seekers in finding the right positions and employers in identifying the most suitable candidates in a complex and dynamic job market. This system has been developed to better meet the needs inadequately addressed by traditional job recommendation systems (JRS), by combining both syntactic and semantic approaches. HJRS consists of three main components: (1) Resume Recommender System Ontology-RRSO, (2) Ontology-Based Resume Text Parsing Module (OntRTPM), and (3) System Database. While RRSO is used to semantically interpret job postings and candidate information, OntRTPM extracts the necessary personal information, educational background, work experiences, and skills from candidates' resumes. The candidate extracted data is saved in the appropriate field of the system database, and an OWL file is created for the candidate. The system operates through four main steps: (1) preprocessing, (2) feature processing, (3) ontology matching and labelling, and (4) saving structured resume data. In the preprocessing stage, data is cleaned and normalized; in the feature processing stage, concept extraction and semantic matching operations are performed. As a result of these processes, candidates' resumes are stored in both OWL file and the relevant tables in the system database. The dataset used consists of 100 anonymized Turkish resumes randomly selected from the database of Kariyer.Net, a large career company in Turkey. OntRTPM has been subjected to accuracy and reliability tests by extracting information from these resumes. The system aims to significantly improve job search and recruitment processes by providing more accurate and effective recommendations for both job seekers and employers in the job market. Future work will include expanding the resume dataset and further optimizing system performance.