Authors:

Jens Dörpinghaus, Johanna Binnewitt, Stefan Winnige

Abstract:

In this paper, we will evaluate if we can build an interoperable knowledge graph on job ads mapping to existing language resources that describe skills utilizing contextual data. Job ads are one way for a company to recruit new employees, but are also widely used in labor market research and analysis. Beside general information about the hiring company and the working conditions, job ads document the current skill needs. For our analysis, we propose a detailed overview of the existing knowledge graph representation of job ads. This knowledge graph provides contextual data. We present a first attempt to evaluate different German Skill resources and their coverage in job ads and their interoperability.

However, our analysis shows that challenges are not limited to technologies used, but are also related to the different scopes of skill taxonomies and the modelling of skills in general. Contextual data might lead to better results in specific situations.

Thus, we hope that further collaborations on skill taxonomies and ontologies might push the frontiers in job ad analysis. In particular, by integrating more contextual information (e.g. professions), by splitting skills into verbs and objects, and through further discussion about possible applications.