Cause Event Representations for Happiness and Surprise

Cause Event Representations for Happiness and Surprise Sophia Yat Mei Lee a, Ying Chen a, and Chu-Ren Huang a,b aDepartment of Chinese and Bilingual Studies, The Hong Kong Polytechnic University, Hong Kong {sophiaym, chenying3176, churenhuang}@gmail.com bInstitute of Linguistics, Academia Sinica, Taipei Abstract. This paper presents a linguistic analysis of emotions by introducing some concrete linguistic rules in identifying the important elements of an emotion, which are the experiencer and the cause. It specifically discusses two primary emotions, happiness and surprise, in Chinese. These rules can be very useful for emotion detection and classification. We also examine the features of a cause event according to its degree of transitivity, namely agentivity, kinesis, and event participation. We find that the emotion classes (happiness vs. surprise) override the emotion verb types (change-of-state verb vs. homogeneous verb) in terms of cause event features. We believe that this study will have some implications for the linguistic account of causal events as well as the theory of emotions in general. We also hope that this work will shed lights in the automatic detection and classification of emotion in language technology. Keywords: emotion, cause event, experiencer, happiness, surprise. 1 Introduction1 Emotions have been widely studied in and across various disciplines. Various linguistic and psychological theories of emotions were proposed previously (e.g. Plutchik 1980, Frijda 1986, Weiss and Cropanozano 1996). They are generally considered the least tangible aspects of human experience, but exert powerful influence upon our thoughts and actions, and even physical appearance and physiological processes occurring within our bodies (Harkins and Wierzbicka 2001). With its complex nature, there have been considerable debates among the theories on the definition as well as the classification of emotions. Nonetheless, most theories agree that an emotion is composed of or stimulated by a cause event and that the cause events are the most tangible component of emotion (Wierzbicka 1996). As no objective criteria have been offered yet to define and classify cause events empirically, this study aims to examine the interaction between emotions and cause events. As for real world applications, the information of who the experiencer of an emotion is and what triggers such an emotion is an essential element for NLP. Such information is particularly important for economic and political evaluations as well as product reviews. However, identifying the experiencer and the causes of an emotion is very challenging in NLP. We believe that a linguistic analysis of cause events may help generalize some useful patterns for automatic emotion detection and classification. In this paper, we present a linguistic analysis of emotions in terms of their experiencers and causes. Causes refer to the explicitly expressed arguments or events that trigger the presence of the corresponding emotions, which are usually expressed by means of verbs, nominalizations, Copyright 2009 by Sophia Yat Mei Lee, Ying Chen, and Chu-Ren Huang 345 23rd Pacific Asia Conference on Language, Information and Computation, pages 345–354 and can sometimes be event nominals (Sauri et al. 2006). We specifically look at two primary emotions, happiness and …

Content truncated to 500 words