Computing Humorous Metaphors

Abstract

It was experimentally showed that humorous texts can be explained using approaches applied to metaphors, such as the salience-imbalance or the domain-interaction approach. It was also demonstrated that humorous metaphors often include a switch between positive and negative emotions. We propose to construct a computer system able to understand and generate such metaphors. Currently we are constructing a metaphor conceptual network, in which links between concepts are calculated accordingly to their roles in metaphor understanding. This will allow the computer to process metaphors. Next, we will adjust the links distance calculation to match the humorous metaphors (to increase the salience-imbalance, as demonstrated in existing research). We will also use an emotiveness-recognition-system to detect emotive associations towards particular phrases, in order to choose pairs with the emotional switch. The system will be evaluated in user-oriented experiments. Acknowledgements: This work was supported by KAKENHI (Project Number: 23-01348)


Back to Table of Contents