| It has been a growing concern among the artificial intelligence community to design methods to integrate and unify knowledge from disparate domains. It has become imperative to design adaptive systems that can react to dynamic environments through common sense reasoning and learning. A hypothesis is put forward suggesting that intelligent systems acquire new concepts through establishing appropriate associations to the knowledge already present in their long-term memories. This paper proposes a knowledge representation scheme that supports the hypothesis by modeling domain independent ontologies in the form of objects and semantic primitives. A consistent representation, across entities from different domains, is built by asking certain logical questions about how each entity performs the following two primitive actions: "to change" and "to exist." The scheme is shown to facilitate analogy based learning, common sense reasoning, knowledge acquisition, configuration, and diagnostics. |
