Concept Learning for EL ++ by Refinement and Reinforcement Mahsa Chitsaz, Kewen Wang, Michael Blumenstein, and Guilin Qi School of Information and Communication Technology, Griffith University, Australia School of Computer Science and Engineering, Southeast University, China mahsa.chitsaz@griffithuni.edu.au, {kewen.wang,m.blumenstein}@griffith.edu.au, gqi@seu.edu.cn Abstract. Ontology construction in OWL is an important and yet time- consuming task even for knowledge engineers and thus a (semi-) automatic approach will greatly assist in constructing ontologies. In this paper, we propose a novel approach to learning concept definitions in EL ++ from a collection of assertions. Our approach is based on both refinement operator in inductive logic programming and reinforcement learning algorithm. The use of reinforcement learning significantly re- duces the search space of candidate concepts. Besides, we present an experimental evaluation of constructing a family ontology. The results show that our approach is competitive with an existing learning system for EL. Keywords: Concept Learning, Description Logic EL ++ , Reinforcement Learning, Refinement Operator. 1 Introduction Description logics have become a formal foundation for ontology languages since the Web OntologyLanguage (OWL), which is adapted as the World Wide Web Consortium (W3C) standard for ontology languages. Recently, OWL has evolved into a new standard OWL 2 1 , which consists of three ontology language pro- files: EL, QL and RL. OWL 2 standard provides different profiles that trade some expressive power for the efficiency of reasoning, and vice versa. Depending on the structure of the ontologies and the reasoning tasks, one can choose either of these profiles. OWL 2 EL, which is based on EL ++ [1], is suitable for appli- cations employing ontologies that contain very large numbers of properties and classes, because the basic reasoning problems can be performed in time that is polynomial with respect to the size of the ontology. The ontology consists of a terminology box, Tbox, and an assertion box, Abox. Besides, concept learning concerns learning a general hypothesis from the given examples of the ontology that we want to learn; those examples are instances 1 http://www.w3.org/TR/owl2-overview/ P. Anthony, M. Ishizuka, and D. Lukose (Eds.): PRICAI 2012, LNAI 7458, pp. 15–26, 2012. c  Springer-Verlag Berlin Heidelberg 2012