MAP: An Abstraction-Based Metaphor Analysis Program for Overcoming Cross-Modal Challenges

Abstract

Much of the research on computational metaphor interpretation is based on metaphors that are prevalent in the language being processed. Moreover, the more detailed work in this area tends to focus primarily on specific domains of discourse. Such approaches, while of interest and value, are not adequate to handle novel metaphoric expressions that occur in a context unrestricted by domain, such as an ordinary conversation or a digression from a domain-specific context. This chapter describes MAP (Metaphor Analysis Program), a computer program that processes both novel and conventional cross-modal metaphor without restriction to a particular domain of discourse. MAP relies on the kind of semantic analysis that models what humans are hypothesized to do when they extend a literal meaning to a metaphoric one, especially when there is no clear indication of a discursive context. To do so, the model depends on an analysis of a metaphorically used word in its literal sense and the role of this sense in structuring the topic of the metaphor. Problems of multiple metaphoric interpretations and the uncertainty of constraints on metaphoric coherence are addressed by a number of vivid illustrations.