OntoJob: Automated Ontology Learning from Labor Market Data

Due to the rapidly changing labor market and the consequently widening information gap between the labor market and education, there is a need for methods that can tackle, or at least ease, the construction of labor market ontologies. The current study set out to examine the viability of Ontology Learning (OL) methods for the (semi-)automated construction of labor market ontologies and / or taxonomies. The purpose of this paper is to propose an unsupervised framework, OntoJob, that can identify and extract from raw vacancy text instances, attributes, and relations, such as job titles, worker qualities, and the non-Taxonomic 'is-A' relations between those concepts, and convert those to an expressive descriptive logic. Evaluation of the extracted worker qualities from OntoJob, using a small sample of 5621 job postings representing 1048 occupations, showed an overall lexical precision of 0.36 and recall of 0.22.