A two-layered Knowledge Architecture for perceptual and linguistic Knowledge

The lack of generality is a structural weakness of knowledge representation formalisms. Here by lack of generality we mean the inability of any given representation to describe the infinite richness and diversity of the world and also its potentially infinite descriptions which are enabled by language. This lack of generality is the main cause of many of the difficulties encountered so far, just think of the problems which have arisen in the effort of creating reusable ontologies. In this thesis we propose a solution to the problem of generality which is based on the key idea that knowledge should not be modeled a priori, at design time, but it should continuously generated, adapted and evolved, from generation to usage. The thesis provides four main contributions: (i) a shared terminology for the characterization of concepts and for their computational representation; (ii) a formalization of the distinction between substance concepts and classification concepts; (iii) the integration of these two notions of concept into a general representation language that organizes them into a hierarchy of increasing abstraction of what is perceived, and (iv) a two-layered knowledge representation formalism, where the first layer allows to represent concepts, as the main devices for achieving generality, and where the second layer allows to represent concepts as the result of “adapting†a description to the current knowledge representation needs and requirements.