Methods and models of automatic ontology construction for specialized domains (case of the Radiation Security)

Abstract

We propose a hybrid, semi-automatic approach that uses the intersection of semantic classes of nouns and verbs built on the domain lexicon and builds kernel ontology from a list of initial concepts and then completes this kernel ontology by new entities detected in a large corpus of texts of international standards of Radiological Safety. The results confirm the important role of initial linguistic modeling and show that the external lexical resources available online can contribute effectively to the resolution of the problem of lexical disambiguation.