JobEdKG: An uncertain knowledge graph-based approach for recommending online courses and predicting in-demand skills based on career choices

Abstract

Modern job markets often require an intricate combination of multi-disciplinary skills or specialist and technical knowledge, even for entry-level positions. Such requirements pose increased pressure on students, new graduates, and job seekers to find suitable sources of information to enhance their employability. Unlike standard competence management systems, this paper presents JobEdKG, which helps job seekers from various backgrounds choose online courses based on their prospective careers in addition to predicting different skills that may be required in their chosen career path. While existing solutions focus on internal institutional data, such as previous student experiences and a fixed set of skills provided by curated datasets, JobEdKG considers external data, recommending online courses that best cover the knowledge and skills required by selected job roles, in addition to extracting and predicting skills of that particular job roles. To achieve this, we first extract skills from job postings and online courses. These skills are linked to job titles, online courses, and other concepts in order to create a knowledge graph (KG). We assign an uncertainty score to each fact in the KG based on the prevalence of the fact in the source data (i.e. job listings and online courses), which results in an uncertain KG (UKG). Finally, we model the constructed UKG in order to infer different relations between the different concepts. The code and the data are available on our GitHub repository (https://github.com/team611/JobEd) and a user interface to browse the KG is available at (http://jobed.datanets.org/).