Abstract:

In today’s dynamic job market, identifying the skills required by employers and helping learners acquire those skills efficiently is a significant challenge. In the context of the ENTEEF (Fostering Entrepreneurship through Freelancing) project, we address the problem of automatically generating structured learning paths using skills extracted from job offers. The ENTEEF project focuses on bridging the gap between education and industry by preparing students (and other target groups) with the competencies needed for freelancing careers. To this end, we propose a methodology that leverages Bayesian Networks (BN) to model relationships between skills extracted from ~30,000 job postings, and we use the network to derive recommended learning sequences. In our approach, each skill is a node in a BN and edges indicate dependency or co-occurrence relationships learned from job market data. We experiment with two weighting methods for the BN’s arcs - mutual information score and normalized mutual information score - to quantify the strength of connections between skills. These weighting strategies, based on information theory, help rank possible skill progressions and guide the generation of optimal learning paths. Our method thus combines data-driven skill extraction with probabilistic modelling to produce learning pathways aligned with real-world job requirements.