Plumeria at SemEval-2022 Task 6: Sarcasm Detection for English and Arabic Using Transformers and Data Augmentation

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 923 - 937 July 14-15, 2022 ©2022 Association for Computational Linguistics Plumeria at SemEval-2022 Task 6: Sarcasm Detection for English and Arabic Using Transformers and Data Augmentation Mosab Shaheen∗ Indian Institute of Technology Kanpur mosab@iitk.ac.in Shubham Kumar Nigam∗ Indian Institute of Technology Kanpur sknigam@iitk.ac.in Abstract The paper describes our submission SemEval-2022 Task 6 on sarcasm detectionto and its five subtasks for English and Arabic. Sarcasm conveys a meaning which contradicts the literal meaning, and it is mainly found on social networks. It has a significant role in un- derstanding the intention of the user. For detect- ing sarcasm, we used deep learning techniques based on transformers due to its success in the field of Natural Language Processing (NLP) without the need for feature engineering. The datasets were taken from tweets. We created new datasets by augmenting with external data or by using word embeddings and repetition of instances. Experiments were done on the datasets with different types of preprocessing because it is crucial in this task. The rank of our team was consistent across four subtasks (fourth rank in three subtasks and sixth rank in one subtask); whereas other teams might be in the top ranks for some subtasks but rank dras- tically less in other subtasks. This implies the robustness and stability of the models and the techniques we used. 1 Introduction Sarcasm is a figurative language where speakers or writers usually mean the contrary of what they say. Recognizing whether a speaker or writer is sarcastic is essential to downstream applications to understand the sentiments, opinions, and beliefs correctly (Ghosh et al., 2020). Sarcasm is ubiqui- tous on the social media text and, due to its nature, can be highly divisive of computational systems that perform tasks on that kind of data such as sen- timent analysis, opinion mining, and harassment detection (Van Hee et al., 2018; Bing, 2012; Rosen- thal et al., 2014; Maynard and Greenwood, 2014). Our team Plumeria participated in SemEval 2022 task 6 (Abu Farha et al., 2022) in all its subtasks on English and Arabic. Previous shared tasks *These authors contributed equally to this work on sarcasm detection (Hee et al., 2018; Ghanem et al., 2019; Ghosh et al., 2020; Abu Farha et al., 2021) have only two subtasks; one is sarcasm detec- tion and another is predicting the type of sarcasm. However, in SemEval 2022 task 6 (Abu Farha et al., 2022) organizers formulate three subtasks for both languages following the methods described in (Oprea and Magdy, 2020). Using the two datasets for English and Arabic, organizers formulate three subtasks as follows: • Subtask A (English and Arabic): It is a bi- nary classification subtask where submitted systems have to predict whether a tweet is sarcastic or not. • Subtask B (for English only): It is a multi- label classification subtask where submitted systems have to predict one or more labels out of six ironic-speech labels: sarcasm, irony, satire, understatement, overstatement, …

Content truncated to 500 words