NeoHire: A Neo4j-Based Intelligent Recruitment System

Abstract

Even though many organisations have worked on their recruitment systems over the years, a considerable part of the hiring process still relies on manually scanning resumes and using keywords to filter candidates they want to hire. All of these methods can help limit basic screening effort, but don’t always identify candidates with the skills and experience to match a job role based on an exact keyword match. With hundreds of applications for a single position, recruiters are expected to shortlist within a limited time, which can be a repetitive, time-consuming, and sometimes inconsistent process. This paper presents a novel recruitment system, called NeoHire, which utilises NLP, ML, and graph database technologies to enhance candidate screening. The system emphasizes the semantics behind resumes and job descriptions over keyword matching by applying the technique of semantic relationships to candidate resumes and job descriptions. The data of the candidate’s resume, including skills, education and work experience, communications and other important details, is extracted by NeoHire from a variety of resume formats and subsequently stored in a Neo4j graph structure. Transformer-based embedding models are employed to assess the contextual similarity between resumes and job requirements, thereby enhancing matching accuracy. The graph-based system also facilitates a better understanding of the connections among candidates, skills, and job roles. The intended system is designed to minimize the manual screening time, enhance screening transparency in the shortlisting process, and offer a scalable recruitment system that fits more closely with the needs of modern recruitment processes.