A Methodology for the Induction of Ontological Knowledge from Semantic Annotations

At the meeting point between machine learning and description logics, we investigate on the induction of structural knowledge from metadata. In the proposed methodology, a basic taxonomy of the primitive concepts and roles is preliminarily extracted from the assertions contained in a knowledge base. Then, in order to deal with the inherent algorithmic complexity that affects induction in structured domains, the ontology is constructed incrementally by refining successive versions of the target concept definitions, expressed in richer languages of the Semantic Web, endowed with well-founded reasoning capabilities.