A corpus-based construction
of emotion verb scales
We examine subclasses of English verbs of emotion, in particular Experiencer verbs like <i>surprise</i>, <i>fear</i> and <i>astonish</i>, which express the causation of gradable emotions. The gradation is richly lexicalized by verbs that denote different degrees of intensity of the same emotion (e.g., <i>surprise, strike, dumbfound, flabbergast</i>). We examine manually constructed groups of verbs expressing different degrees of intensity of one underlying emotion in the light of corpus (Web) data. We employ lexical-semantic patterns that were previously identified as discriminating between the members of pairs of scalar adjectives and apply them to verbs that belong to three broad emotion verbs classes. The results show, first, that the chosen verbs indeed possess scalar qualities; second, they confirm the prior assignment of the verbs into broad classes based on a common underlying emotion; finally, the Web data allow us to construct consistent scales with verbs ordered according to the intensity of the emotion. Future extensions to additional verb classes will lead to a more subtle analysis of the semantics of emotion verbs and their improved representation in lexical resources with potential benefits for translation, language pedagogy and automatic inferencing.