Subjective emotions vs. verbalizable emotions in web texts

Abstract

Cognition and emotions are inseparable. Still, it is not clear to which extent emotions can be characterized by words and how much of emotional feelings are non-verbalizable. Here we approach this topic by comparing the structure of the emotional space as revealed by word contexts to that in subjective judgments, as studied in the past. The number of independent emotions and categories of emotions is a key characteristic of the emotional space. Past research were based exclusively on perceived subjective similarities by participants of experiments. Here we propose and examine a new ap-proach, the similarities between emotion names are obtained by comparing the contexts in which they appear in texts re-trieved from the World Wide Web. The developed procedure measures a similarity matrix among emotional names as dot products in a linear vector space of contexts. This matrix was then explored using Multidimensional Scaling and Hierar-chical Clustering. Our main findings, namely, the underlying dimension of the emotion space and the categories of emotion names, were consistent with those based on subjective judgments. We conclude that a significant part of emotional expe-riences is verbalizable. Future directions are discussed.