| Author: | Lau, Yan Ping Helena |
| Title: | Decoding emotions in social media : a linguistic analysis of implicit emotions and events |
| Advisors: | Lee, Yat Mei Sophia (CBS) Huang, Chu-ren (CBS) |
| Degree: | Ph.D. |
| Year: | 2021 |
| Subject: | Social media Online social networks Discourse analysis Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Chinese and Bilingual Studies |
| Pages: | xiii, 242 pages : color illustrations |
| Language: | English |
| Abstract: | This thesis aims to discover meanings expressed in text by focusing on implicit emotion and to investigate the interaction between implicit emotions and event types. Over the past few decades, the web has revolutionized how information is stored, published, and transmitted. Social media in particular has become one of the most influential communication tools, connecting billions of people around the world in a global communications network which is constantly evolving. This allows new information to spread faster and farther to a wider audience, revealing with greater permanence and clarity to the observer the kinds of emotions that are triggered in individuals through the implicit nature of their response to different events. Previous attempts at emotion analysis have focused mainly on the examination of explicit emotions, either in terms of linguistic syntactic and semantic characteristics or their identification and classification within the field of natural language processing. Explicit emotion refers to the emotion-related information denoted directly by the presence of emotion keywords, such as HAPPINESS, ANGER, and SURPRISE. In this thesis, I focus on an important, yet underdeveloped, branch of emotion analysis, implicit emotion. Implicit emotion refers to the presence of emotion-related information conveyed through inference or connotation instead of emotion keywords. I argue that an in-depth analysis of implicit emotion is a necessary component of emotion analysis. As corpus data has shown, the majority of emotions expressed are implicit in nature and there is a clear gap in existing emotion research, which this current work aims to address. By exploring implicit emotions in responses to different events posted on an online social media website, I attempt to address the following questions: How are such implicit emotions expressed in text? What kinds of events trigger different implicit emotions? In this study, an annotated Chinese event-comment corpus retrieved from Sina Weibo is constructed. With the empirical data gathered from the corpus, a comprehensive analysis on emotion expressions is carried out at the semantic, syntactic and discourse levels. Drawing on the insight of Pavlenko (2008), emotions expressed at a word level are studied in terms of the use of emotion words, emotion-related words, and emotion-laden words. An emotion expressed at the semantic level, whether implicit or explicit, can be identified with ease, by modifying an emotion taxonomy and proposing a list of emojis, emotion-related words and emotion-laden words. I also study the syntactic structures that are frequently used to convey emotions. I claim that words of different parts-of-speech can serve as a good emotion indicator when there are no other linguistic clues found in text. The same word formed in different syntactic structures may express different emotions. Findings show that rhetorical questions are a relatively productive means applied in emotion expressions. At the discourse level, the atypical use of emojis is examined. When the emotion expressed in text and the emotion denoted by the emoji are at odds, the overall emotion is determined primarily by the one expressed in text. Apart from the linguistic features of implicit emotion, the correlation between emotions and events are studied. In lieu of using existing event type classification models, I make use of language resources including TimeML (Sauri et al. 2009), WordNet (Miller 1995) and FrameNet (Baker et al. bamv1998) for the markup of events, event classification and the annotation of frame elements, respectively. Based on the annotated data, I summarize a list of event types which show a preference to a particular emotion and are statistically significant. Furthermore, I also investigate the interplay between emotion, event and semantic role. I confirm the hypothesis that HAPPINESS and ANGER is generally evoked by doers of events associated with that emotion. I also conclude that SADNESS is sometimes elicited by undergoers of events associated with SADNESS, and sometimes triggered by doers of events when the situation leaves the doers no option. The linguistic account of implicit emotion directly helps to paint a fuller picture of the forms and representations of implicit emotions. First of all, a Chinese event-comment corpus is constructed, which provides valuable resources for emotion studies from the linguistic and computational perspectives. Second, the proposed linguistic cues and the syntactic structures may serve as the features for computational models and classifiers. This thesis aims to shed light not only on the inference and identification of implicit information, but also on the automatic classification and detection of implicit emotions. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/10820