Causing Emotion in Collocation:An Exploratory Data Analysis

Causing Emotion in Collocation: An Exploratory Data Analysis Pei-Yu Lu, Yu-Yun Chang and Shu-Kai Hsieh Graduate Institute of Linguistics, National Taiwan University No. 1, Sec. 4, Roosevelt Road, Taipei, Taiwan, 10617 {emily.lulala, yuyun.unita, shukai} @gmail.com Abstract This paper aims to seek approaches in investigating the relationships within emotion words under linguistic aspect, rather than figuring out new algorithms or so in processing emotion detection. It is noted that emotion words could be categorized into two groups: emotion-inducing words and emotion-describing words, and emotion-inducing words would be able to trigger emotions expressed via emotion-describing words. Hence, this paper takes the social network Plurk, the emotion words are from the study on Standard Stimuli and Normative Responses of Emotions (SSNRE) in Taiwan and the National Taiwan University Sentiment Dictionary (NTUSD) as corpus, combining with Principle Component Analysis (PCA) and followed collocation approach, in order to make a preliminary exploration in observing the interactions between emotion-inducing and emotion-describing words. From the results, it is found that though the retrieved Plurk posts containing emotion-inducing words, polarities of the induced emotion-describing words contained within the posts are not consistent. In addition, the polarities of posts would not only be influenced by emotion words, but negation words, modal words and certain content words within context. Keywords: sentiment analysis, emotion word, collocation. 1. Introduction Sentiment analysis has recently become a prevalent trend in the field of natural language processing, and has wide applications for industry, policy making, sociology, psychology and so on. Various approaches have been proposed with impressive experimental or computational evidence, from document-level analysis to sentence-level or even phrasal-level analysis [1]. Among most studies, the Sentiment/Emotion-labeled Lexicon is taken as an indispensible lexical resource for the improvement of emotion classification accuracy. However, by assuming the static correspondence of word-emotion, most studies have neglected the fact that emotion words are not fixed with specific valence but are influenced under diverse contexts. On account of contextual effects of emotion, [2-4] have firstly introduced a notion inspired by cognitive linguistics - emotion cause event - that refers to “the explicitly expressed arguments or events that trigger the presence of the corresponding emotions.” A set of Proceedings of the Twenty-Fifth Conference on Computational Linguistics and Speech Processing (ROCLING 2013) 236 linguistic cues is proposed to detect the cause events, resulting a valuable corpus resource for the task of emotion classification. Despite that there are some explicit causers that might trigger emotions via context, a recent large-scaled interdisciplinary emotion research project [5, 6] has focused on the emotion words and found that they do help capture the emotion perceptions [7], and can thus be employed in emotion-related processing tasks. As designed in [5, 6] emotion words can be further grouped into emotion-inducing (情緒誘發詞) emotion-describing words (情緒描述 詞 ). Emotions are mainly divided into two polarities: positive and negative. Emotion-inducing words encode the underlying repository knowledge to be able to elicit emotion-describing words. Therefore, in this study, we assume that the emotion-inducing word can be treated as the pivot in emotion …

Content truncated to 500 words