Automatic ontology construction based on clustering nucleus

Ontology construction is the core task of ontology-based knowledge representation. This paper explores a semantic description approach based on primitive structure, which benefits ontological relation description in a more precise and concrete way. In view of primitive structure, this paper introduces an approach to extract primitive structures of words based on a multi-label learning model, correlated label propagation. Also, this paper proposes an approach to recognize clustering nucleuses in word clusters heuristically. By this approach, more precise ontological relations are able to be discovered automatically.