Affective Computing and Sentiment Analysis (eBook)

Acknowledgments 6
Introduction: Affect Computing and SentimentAnalysis 7
References 10
Contents 11
Contributors 13
1 Understanding Metaphors: The Paradox of Unlike Things Compared 15
1.1 Introduction 15
1.2 The Metaphor Paraphrase Problem and the Priority of the Literal 16
1.3 Understanding Metaphors: Comparison or Categorization? 17
1.4 How Novel Categories Can Be Named: Dual Reference 18
1.5 Understanding Metaphors and Similes 20
1.6 The Metaphor Paraphrase Problem Revisited 22
1.7 Comparison Versus Categorization Revisited 23
1.8 Conclusions 24
References 25
2 Metaphor as Resource for the Conceptualisation and Expression of Emotion 27
2.1 Background 27
2.2 Metaphorical Conceptualisation of Emotions in English 28
2.2.1 Conceptualisation of Emotion 28
2.2.2 Description and Expression of Emotion 30
2.3 Contribution of English Metaphor Themes to the Expression of Emotion 31
2.3.1 Metalude Data for Evaluation 31
2.3.2 Evaluative Transfer 32
2.3.3 Evaluation Dependent on Larger Schemata 34
2.3.4 Ideology and Evaluation 35
2.3.5 The Role of Multivalency and Opposition in Metaphor Themes 37
2.4 Conclusion 39
References 39
3 The Deep Lexical Semantics of Emotions 40
3.1 Introduction 40
3.2 Identifying the Core Emotion Words 41
3.3 Filling Out the Lexicon of Emotion 41
3.4 Some Core Theories 43
3.5 The Theory and Lexical Semantics of Emotion 44
3.6 Summary 46
References 47
4 Genericity and Metaphoricity Both Involve Sense Modulation 48
4.1 Background 48
4.2 Dynamics of First-Order Information 51
4.2.1 Some Intuitions About Revision 51
4.2.2 A Formal Model of First-Order Belief Revision 52
4.2.3 First-Order Belief Revision Adapted to Sense Extension 53
4.3 Ramifications for Metaphoricity 55
4.4 Metaphoricity and Genericity 57
4.5 Particulars of the Class-Inclusion Framework 60
4.6 Final Remarks 63
References 63
5 Affect Transfer by Metaphor for an Intelligent Conversational Agent 65
5.1 Introduction 65
5.2 Affect via Metaphor in an ICA 67
5.3 Metaphor Processing 68
5.3.1 The Recognition Component 68
5.3.2 The Analysis Component 70
5.4 Examples of the Course of Processing 73
5.4.1 You Piglet 73
5.4.2 Lisa Is an Angel 74
5.4.3 Mayid Is a Rock 74
5.4.4 Other Examples 74
5.5 Results 75
5.6 Conclusions and Further Work 76
References 77
6 Detecting Uncertainty in Spoken Dialogues: An Exploratory Research for the Automatic Detection of Speaker Uncertainty by Using Prosodic Markers 79
6.1 Introduction 79
6.2 Related Work 79
6.2.1 Defining (Un)certainty 79
6.2.2 Linguistic Pointers to Uncertainty 81
6.2.3 Prosodic Markers of Uncertainty 81
6.3 Problem Statement 82
6.4 Data Selection 83
6.4.1 Selection of Meetings 83
6.4.2 Data Preparation and Selection 83
6.4.3 Statistical Analysis 84
6.5 Experimentation 85
6.5.1 Hedges --vs-- No Hedges 85
6.5.2 Uncertain Hedges --vs-- Certain Hedges 86
6.5.3 Distribution of Hedges Over Dialogue Acts 88
6.6 Conclusions 88
References 89
7 Metaphors and Metaphor-Like Processes Across Languages: Notes on English and Italian Language of Economics 90
7.1 Introduction 90
7.2 Corpus and Method 92
7.2.1 Corpus 92
7.2.2 Method 92
7.3 Analysis 93
7.3.1 Constitutive Metaphors 93
7.3.2 Pedagogic Metaphors 96
7.3.3 Universal vs. Culture-Specific Metaphors 96
7.4 Conclusion 97
References 98
8 The `Return' and `Volatility' of Sentiments: An Attempt to Quantify the Behaviour of the Markets? 100
8.1 Introduction 100
8.2 Metaphors of `Return' and of `Volatility' 100
8.3 The Roots of Computational Sentiment Analysis 103
8.4 A Corpus-Based Study of Sentiments, Terminology and Ontology Over Time 104
8.4.1 Corpus Preparation and Composition 105
8.4.2 Candidate Terminology and Ontology 105
8.4.3 Historical Volatility in Our Corpus 106
8.5 Afterword 108
References 109
9 Sentiment Analysis Using Automatically Labelled Financial News Items 111
9.1 Introduction 111
9.2 Data and Method 112
9.2.1 Training and Testing Corpus 112
9.2.2 Feature Types 112
9.2.3 Feature Selection and Counting Methods 113
9.2.4 News Items and Stock Price Correlation 114
9.2.5 Feature Selection and Semantic Relatedness of Documents 115
9.3 Results 116
9.3.1 Horizon Effect 116
9.3.2 Polarity Effect 117
9.3.3 Range Effect 118
9.3.4 Effect of Adding a Neutral Class on Non-cotemporaneous Prices: One- and Two-Days Ahead 118
9.3.5 Conflating Two Classes 119
9.3.6 Positive and Negative Features 120
9.4 Discussion 121
9.4.1 Lack of Independent Testing Corpus 121
9.4.2 Pool of Features 122
9.4.3 Size of Documents 122
9.4.4 Trading Costs 122
9.5 Conclusion and Future Work 122
References 123
10 Co-Word Analysis for Assessing Consumer Associations: A Case Study in Market Research 125
10.1 Introduction 125
10.2 Conceptual Background 126
10.2.1 Consumer Associations and Mental Processing 126
10.2.2 Drawbacks of Manual Data Analysis 127
10.2.3 Requirements for Automated Co-Word Analysis 127
10.3 Technique and Implementation 128
10.3.1 Import of Text Sources 129
10.3.2 Processing of Text 129
10.3.3 Graph Creation and Clustering 129
10.4 Exemplary Case Study 130
10.5 Conclusion and Outlook 132
References 133
11 Automating Opinion Analysis in Film Reviews: The Case of Statistic Versus Linguistic Approach 135
11.1 Introduction 135
11.2 Related Work 136
11.2.1 Machine Learning for Opinion Analysis 136
11.2.2 Linguistic Methods of Opinion Analysis 137
11.3 Linguistic and Machine Learning Methods: A Comparative Study 140
11.3.1 Linguistic Approach 140
11.3.2 Machine Learning Approach 143
11.4 Conclusion and Prospects 147
References 148
Afterword: `The Fire Sermon' 151
Name Index 155
Subject Index 157