TOWARD COMPUTATIONAL MODEL OF UNDERSTANDING METAPHORS

This paper presents a computational method of calculating the measure of salience in understanding metaphors. We mainly treat metaphors in the form of “A is (like) B,” in which “A” is called target concept , and “B” is calledsource concept . In understanding a metaphor, some properties of the source concept are transferred to the target concept. In the transfer process, we first have to select the properties of the source concept that can be more preferably transferred to the target concept. The measure of salience represents how typical or prominent the property is and is used to measure the transferability of the property. By introducing the measure of salience, we have to consider only the high salient properties after the selection. The measure of salience was calculated from Smith & Medin’s probabilistic concept according to Tversky’s two factors; intensity and diagnostic factor.