An Ontology Enrichment Approach by Using DBpedia

Abstract

Over the past decade, an increasing number of methods have been proposed for (semi-) automatic generation of ontology from text. However, the ontology generated by these methods usually does not meet the needs of many reasoning-based applications in different domains since most of these methods aim at generating inexpressive ontologies e.g. bare taxonomies. In this paper, a new ontology enrichment approach is proposed in which Web of Linked Data (in particular, DBpedia as one of the huge Linked Data datasets) is used as background knowledge beside text in order to recognize new ontological relations, specifically object properties, for ontology enrichment. In other words, this enrichment approach can be considered as a post-processing step for the "Relations" layer (i.e. the fifth layer) in Ontology Learning Stack, aiming at recommending new object properties to the ontology engineers enabling them to create much more expressive ontologies. This is actually a complementary approach to our recent approach towards adding Linked Data to ontology learning layers where we aimed at improving the functions associated to the "Synonyms" layer, the "Concept Formation" layer and the "Concept Hierarchy" layer of ontology learning stack. In order to evaluate the approach, a customized experimental design is introduced called the "Pseudo Gold Standard based Ontology Evaluation" in which the results obtained by a human expert are compared against those obtained automatically. Finally, the experimental results showed a satisfactory improvement in learning object properties.