EXTENSIONAL SEMANTIC NETWORKS: THEIR REPRESENTATION, APPLICATION, AND GENERATION

ABSTRACT This chapter introduces extensional semantic networks (ESNs), a new kind of semantic network that allows two different levels of interpretation, namely, the intensional and the extensional. We present the theoretical background on which these networks have been designed. In addition, we give both precise definitions and practical examples of ESNs and, finally, are concerned with application areas for and the automatic generation of the networks. In an introductory section, we investigate the relationship between concepts and their extensions, i.e., the objects denoted by these concepts; a set of semantic relations that hold between concepts is introduced and some of their properties are derived. Section 2 contains a formal definition of the hierarchical ESNs that are constructed from concepts and relations. Furthermore, the two interpretations of an ESN, on the conceptual level and on the object level, are developed, including a treatment of negation and quantification. Section 3 is concerned with the use of these networks in information retrieval systems and demonstrates the appropriateness of their application in question-answering systems. In Section 4, we discuss how ESNs can be generated automatically from natural language discourse, emphasis being placed on a transparent description of the stepwise generation of a network.