Learning of Semantic Relations between Ontology Concepts using Statistical Techniques

Acquiring domain knowledge for constructing ontologies is a resource demanding and time consuming task. Thus, the automatic or semi-automatic construction, enrichment and adaptation of ontologies, the so-called ontology learning task is highly desired. Although an emerging field, a significant amount of research has been performed in ontology learning, leading to a large number of proposed approaches and practical systems. This paper presents our approach on automated learning of ontologies from texts which are semantically annotated with instances of ontologies’ concepts. Statistical techniques are applied to metadata extracted from the annotated texts, to discover semantic relations among the annotated concepts as well as to find cardinality restrictions for these concepts and their relations. The proposed method was applied to corpora from two different domains, athletics and biomedical, and was evaluated against the existing manually created ontologies for these domains.