Node Subsumption in a Propositional Semantic Network with Structured Variables

We present a formal description of a propositional semantic-network-based knowledge representation system. Variables in this representation are not atomic and have potentially complex structure. We start from the individual components of a semantic network system, atomics nodes and relations that connect nodes, and provide a complete speci cation for the structure of nodes and a subsumption procedure between nodes. We di er from other work in subsumption in that the representation language is uniform and based on a rst-order predicate logic. There is no distinction between taxonomic and axiomatic knowledge. The utility of the subsumption procedure for processing natural language is illustrated.