Abstract: In this paper, we describe a CV recommender system with a focus on two properties. The first property is the ability to classify candidates into roles based on automatic processing of their CV documents. The second property is the ability to recommend skills to a candidate which are not listed in their CV, but the candidate is likely to have them. Both features are based on skills extraction from a textual CV document. A spectral skill clustering is precomputed for the purpose of candidate classification, while skill recommendation is based on various similarity-based strategies. Experimental results include both automatic experiments and an empirical study, both of which demonstrate the effectiveness of the presented methods.
Authors: Adrian S Kurdija, Petar Afric and Lucija Šikić (University of Zagreb, Faculty of Electrical Engineering and Computing, Croatia); Boris Plejić (Ericsson Nikola Tesla, Zagreb, Croatia); Marin Šilić (University of Zagreb, Croatia); Goran Delac, Klemo Vladimir and Sinisa Srbljic (University of Zagreb, Faculty of Electrical Engineering and Computing, Croatia)
Email: adrian.kurdija@fer.hr, petar.afric@fer.hr, lucija.sikic@fer.hr, boris.plejic@ericsson.com, marin.silic@fer.hr, goran.delac@fer.hr, klemo.vladimir@fer.hr, sinisa.srbljic@fer.hr