JobHive: A Semantic Path-Based Platform for E-Recruitment Recommendation

This paper presents JobHive, a recommender system based on knowledge graphs to provide improved recommendations by aligning candidate resumes with job requirements, considering both explicit and implicit skills. By integrating semantic similarity computations, the system ensures comprehensive match quality for job seekers and employers. The matching algorithm calculates a similarity score between job offers and resumes by comparing skills, experience, and inferred skills. It uses a Transformer-based Sequential Denoising Auto-Encoder (TSDAE) for contextualized understanding, which generates comprehensive representations of entities to improve semantic similarity assessments. Additionally, the algorithm uses a knowledge graph to understand connections between entities, allowing it to find the best matches by considering both direct and indirect relationships. The evaluation of the matching algorithm for JobHive demonstrated its effectiveness in ranking resumes according to job offers. Tested with 40 job offers and 240 resumes, the algorithm achieved high relevance scores, indicating it closely matched manual rankings.