Automatic Detection of Irony

Jihen Karoui is Research and Development Project Manager at AUSY, France. Farah Benamara is a Senior Lecturer at Paul Sabatier University in Toulouse, France. Véronique Moriceau is a Senior Lecturer at Paul Sabatier University.

• Preface ixIntroduction xiChapter 1. From Opinion Analysis to Figurative Language Treatment 11.1. Introduction 11.2. Defining the notion of opinion 31.2.1. The many faces of opinion 31.2.2. Opinion as a structured model 41.2.3. Opinion extraction: principal approaches 51.3. Limitations of opinion analysis systems 71.3.1. Opinion operators 81.3.2. Domain dependency 91.3.3. Implicit opinions 101.3.4. Opinions and discursive context above phrase level 111.3.5. Presence of figurative expressions 121.4. Definition of figurative language 131.4.1. Irony 131.4.2. Sarcasm 181.4.3. Satire 201.4.4. Metaphor 211.4.5. Humor 221.5. Figurative language: a challenge for NLP 231.6. Conclusion 23Chapter 2. Toward Automatic Detection of Figurative Language 252.1. Introduction 252.2. The main corpora used for figurative language 272.2.1. Corpora annotated for irony/sarcasm 282.2.2. Corpus annotated for metaphors 332.3. Automatic detection of irony, sarcasm and satire 362.3.1. Surface and semantic approaches 362.3.2. Pragmatic approaches 392.4. Automatic detection of metaphor 512.4.1. Surface and semantic approaches 522.4.2. Pragmatic approaches 532.5. Automatic detection of comparison 582.6. Automatic detection of humor 582.7. Conclusion 61Chapter 3. A Multilevel Scheme for Irony Annotation in Social Network Content 633.1. Introduction 633.2. The FrIC 653.3. Multilevel annotation scheme 663.3.1. Methodology 663.3.2. Annotation scheme 693.4. The annotation campaign 793.4.1. Glozz 793.4.2. Data preparation 803.4.3. Annotation procedure 813.5. Results of the annotation campaign 833.5.1. Qualitative results 833.5.2. Quantitative results 843.5.3. Correlation between different levels of the annotation scheme 893.6. Conclusion 93Chapter 4. Three Models for Automatic Irony Detection 954.1. Introduction 954.2. The FrICAuto corpus 974.3. The SurfSystem model: irony detection based on surface features 994.3.1. Selected features 994.3.2. Experiments and results 1014.4. The PragSystem model: irony detection based on internal contextual features 1044.4.1. Selected features 1044.4.2. Experiments and results 1094.4.3. Discussion 1164.5. The QuerySystem model: developing a pragmatic contextual approach for automatic irony detection 1184.5.1. Proposed approach 1184.5.2. Experiments and results 1224.5.3. Evaluation of the query-based method 1234.6. Conclusion 124Chapter 5. Towards a Multilingual System for Automatic Irony Detection 1275.1. Introduction 1275.2. Irony in Indo-European languages 1285.2.1. Corpora 1285.2.2. Results of the annotation process 1305.2.3. Summary 1395.3. Irony in Semitic languages 1405.3.1. Specificities of Arabic 1425.3.2. Corpus and resources 1435.3.3. Automatic detection of irony in Arabic tweets 1465.4. Conclusion 149Conclusion 151Appendix 155References 169Index 189