Study on emotion spaces with centrality measure

Abstract

In this paper, we study existing models of emotion space using centrality, which is borrowed from network theory, to identify key emotions as the central nodes in a network, for the purposes of understanding the existing emotion spaces better in a new way. With several different definitions of centrality, key emotions are identified for four existing emotion space models. We also propose a method for integrating existing spaces to build a refined space with more emotion terms. Each model identified different key emotions. When we reduced emotion spaces such that they each contained 21 common emotions, the key emotions identified remained different, implying fundamental structural differences among existing emotion space models. These findings call for further experimental verification and the refinement of emotion models for future research to make it more useful in emotion research.