Primitive Matching Approach : A Method for Eliciting Metaphorical Knowledge about Viewing

Abstract

Cross-modal metaphors are based on some "view-relations" between concepts in different domains, i.e. knowledge about viewing a mental (or abstract) concept as another Physical concept. Our goal is to establish an effective method of computationally eliciting those pieces of knowledge from cross-modal metaphors we human beings use in everyday life. Since we understand that analogical mappings between two non-identical concepts underlying cross-modal metaphors represent viewing-relations, we proposed a method for determining analogical mappings of metaphors as follows. Basically, we employ the structural consistency constraint in determining analogical mappings which allows the mappings of non-identical two relational predicates. In general, however, only this simple version of structural consistency is too weak as a constraint and tends to produce so many meaningless alternatives of analogical mappings. This paper addresses the problem by employing a semantic constraint, celled primitive matching, only at the initial stage of finding analogical mapping ; based on the representation that verbal concepts are described by use of abstract primitives only, the program first searches for the pairs of first-order literals with exactly the same Primitive Predicates which yield some initial mappings. And after that, the program searches for the whole correspondence between two concepts which does not contradict the initial mappings, based on the simple version of structure consistency constraint. In this paper, we made an experiment of analizing some cross-modal verbal metaphors by the primitive matching approach, as a result of which we show that the approach proposed is more advantageous in cpu-time saving than the simple version of structural consistency approach in some classes of verbal metaphors, and we also illustrate the obtained knowledge about viewing mental concepts as physical ones, which is general enough for use in interpreting other metaphoric expressions.