Evaluating Artificial Intelligence-based Job–Skill Matching for Early-Career Entrants: A Review

Abstract

A proposed and described architecture of an artificial intelligence-based job-skill matching system is suggested in this paper in response to the highly high employability gap experienced by university students, especially within the structural graduate unemployment problem of Sri Lanka. The system is based on an advanced pipeline comprising of natural language processing and deep learning models (fine-tuned BERT/SkillBERT) to extract skills in student curriculum vitae (CVs) and industry job descriptions ( JDs) in a robust manner. The main innovation is the skill gap generator (The "Bridge"), which is used to measure the deficiency in terms of semantic cosine similarity of skill embeddings and produces personalized and prioritized gap reports. Targeted learning paths are then offered by a hybrid recommendation engine. It is assumed that this solution will greatly enhance student career preparedness and will make university curriculums in line with current IT and Business industry needs.