Exploring Emotional Words for Chinese Document Chief Emotion Analysis

Exploring Exploring Exploring Exploring Emotional Emotional Emotional Emotional Words Words Words Words for for for for Chinese Chinese Chinese Chinese Document Document Document Document Chief Chief Chief Chief Emotion Emotion Emotion Emotion Analysis Analysis Analysis Analysis∗∗∗∗ Yunong Wu, Kenji Kita, Fuji Ren, Kazuyuki Matsumoto and Xin Kang Faculty of Engineering, University of Tokushima, Minami-josanjima, Tokushima, 771-0206, Japan wuyunong@iss.tokushima-u.ac.jp {kita, ren, matumoto, kangxin}@is.tokushima-u.ac.jp Abstract. Abstract. Abstract. Abstract. In this study, we develop a document emotion analysis model by making use of the function of emotional words annotation. Eight basic emotion categories have been selected, including Expect, Joy, Love, Surprise, Anxiety, Sorrow, Anger and Hate, for both the word and the document level emotion analysis. We introduce two parameters, term relevance and term frequency, to evaluate the relations between the word emotions and the document emotions. Promising experimental results reveal the effectiveness of our document emotion analysis model under different emotion situations. Keywords: Keywords: Keywords: Keywords: emotional words annotation, document emotion analysis, term relevance, term frequency. ∗ This work was supported in part by a grant from the Grant-in-Aid for Scientific Research (B) numbered 21300036 and Young Scientists (B) numbered 23700252 from the Japan Society for the Promotion of Science, and Grant-in-Aid for Challenging Exploratory Research numbered 21650030 from the Ministry of Education, Science, Sports and Culture. Copyright 2011 by Yunong Wu, Kenji Kita, Fuji Ren, Kazuyuki Matsumoto and Xin Kang 25th Pacific Asia Conference on Language, Information and Computation, pages 597–606 597 1111 Introduction Introduction Introduction Introduction Document emotion analysis has become a popular subject in natural language processing studies, which however has been reported to be a challenging problem since people's inner emotions are not easy to be represented exactly only through semantic labels given by machines. Generally speaking, emotion analysis can be divided into two quite different levels: the coarse emotion analysis and the fine-grained emotion analysis. For the former study, researchers have conducted much work on dividing the documents into three classes according to their polarities: the positive documents, the negative documents, and the neutral documents. While for the latter study, researchers often concentrate on extracting some significant types of emotions, such as Expect, Joy, Love, Surprise, Anxiety, Sorrow, Anger and Hate. And this extracting process covers different textual levels, including words, sentences, paragraphs as well as documents. In this paper we are focusing on two important parts of the fine-grained emotion analysis: emotional words annotation and document emotion analysis. Totally eight emotion categories are selected, namely, Expect, Joy, Love, Surprise, Anxiety, Sorrow, Anger and Hate. Under the subject of emotional words annotation, most studies are concentrating on exploring a variety of emotion lexicons. Such studies include (Hu and Liu, 2004; Tokuhisa et al., 2008), in which machine learning techniques are also utilized. However, at most of the time, using emotion lexicons could not be a good idea for analyzing the emotions in the real-world texts, since the emotion lexicons are static while in the real-world the word emotions could always change in different contexts. When a certain …

Content truncated to 500 words