AI-driven semantic similarity-based job matching framework for recruitment systems

Abstract

This paper presents a real-time online recruitment application that integrates semantic similarity and artificial intelligence (AI) to improve job-candidate matching. It addresses the growing volume of job applications and the limitations of traditional keyword-based systems, which often fail to capture contextual meaning and complex semantic relationships in job-candidate alignment. The proposed system leverages natural language processing (NLP) techniques, specifically TF-IDF vectorization, cosine similarity scoring, and domain-specific keyword weighting, to interpret conceptual relevance between resumes and job descriptions, enabling more accurate and inclusive recruitment outcomes. This research developed the system in Python and evaluated it using simulated and real-world recruitment datasets. Experimental results show that the semantic model consistently outperforms keyword-based matching across diverse job domains. For instance, in simulated tests, similarity scores reached 0.74 in the Software Engineer domain, compared to just 0.35 using keyword-based methods. Real-world evaluations further confirmed the model’s effectiveness, with semantic scores of 0.83, 0.76, and 0.74 for the Hadoop, Data Science, and PMP domains, respectively. In contrast, the corresponding keyword-based scores remained below 0.17. Additionally, the system performs well in aligning generalist and specialist profiles, achieving a score of 0.88 for Data analysis roles. These findings validate the system’s robustness, scalability, and ability to interpret varied terminology across job sectors. The research presents a scalable, AI-driven framework that supports context-aware, fair, and accurate job matching, significantly advancing intelligent recruitment technology.