Towards the Natural Ontology of Wikipedia
Andrea Giovanni Nuzzolese, Aldo Gangemi, Valentina Presutti, Paolo Ciancarini · 2013
Abstract. In this paper we present preliminary results on the extraction of ORA: the Natural Ontology of Wikipedia. ORA 4 is obtained through an automatic process that analyses the natural language definitions of DBpedia entities provided by their Wikipedia pages. Hence, this ontology reflects the richness of terms used and agreed by the crowds, and can be updated periodically according to the evolution of Wikipedia. 1 An ontology for Wikipedia The DBpedia Ontology 5 (DBPO) and Yago [7] are the two reference ontologies for DBpedia. Both of them provide only partial extensional and intensional coverage of DBpedia entities because they rely on Wikipedia categories (Yago) and infoboxes (DBpedia), which induce an intrinsic limit of domain coverage [3]. Tìpalo [3] is a tool that automatically produces a RDF taxonomy of types (aligned with WordNet and Dolce) for a DBpedia entity by analysing its natural language definition in Wikipedia. This approach is aimed at identifying the most natural types for an entity as they are expressed by the crowds. By running Tìpalo on the whole DBpedia we aimed at deriving a natural ontology for Wikipedia and approximating as much as possible a complete domain coverage. In this paper, we show the results obtained so far from this process: the first version of the Natural Ontology of Wikipedia (ORA) 6, and we discuss emerging issues and possible solutions for its refinement. This article is organized as