An Integrated Model of Lexico-Semantic and Conceptual Processing for the Treatment of Natural Language Metaphors

Research into the computational properties of the human metaphor faculty can be categorized according to two different perspectives: the lexico-semantic approach, most favoured by linguists, which sees metaphor as a predominantly linguistic phenomenon and thus attempts to model it as such; and the conceptual approach, which sees metaphor as a deep-seated cognitive agency that systematically organizes much, if not all, of our conceptual architecture. Both of these perspectives have brought different algorithmic and representational apparatus to bear on the task of metaphor interpretation and generation, but each is incomplete in itself. The lexico-semantic approach provides a good account of how different word combinations can give rise to creative figurality, while the conceptual approach explicates the deep structural mechanics of metaphor and analogy, connecting both into the breadth of world knowledge. This paper presents an integrated view of metaphor that unifies the lexico-semantic and conceptual approaches, explaining how metaphor is governed both by word and world knowledge. Emerging from this synthesis is a new role for metaphor in the workings of case grammars and the issue of case / filler compatibility.