I2C at SemEval-2022 Task 6: Intended Sarcasm in English using Deep Learning Techniques

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 856 - 861 July 14-15, 2022 ©2022 Association for Computational Linguistics Abstract Sarcasm is often expressed through several verbal and non-verbal cues, e.g., a change of tone, overemphasis in a word, a drawn- out syllable, or a straight looking face. Most of the recent work in sarcasm detection has been carried out on textual data. This paper describes how the problem proposed in Task 6: Intended Sarcasm Detection in English (Abu Arfa et al. 2022) has been solved. Specifically, we participated in Subtask B: a binary multi-label classification task, where it is necessary to determine whether a tweet belongs to an ironic speech category, if any. Several approaches (classic machine learning and deep learning algorithms) were developed. The final submission consisted of a BERT based model and a macro-F1 score of 0.0699 was obtained. 1 Introduction Existing social media analysis systems are limited by their inability to accurately detect and interpret figurative language. Sarcasm is often used by individuals to express opinions on complex matters and regarding specific targets (Carvalho et al. 2009). Early computational models for verbal and irony and sarcasm detection have relied on shallow methods exploiting conditional token count regularities. But lexical clues alone are insufficient to discern sarcasm intent. Appreciating the context of expression is critical for this; even for humans (Wallace et al. 2014). Indeed, the exact same sentence can be interpreted as literal or sarcastic, depending on the speaker. Consider the sarcastic tweet in Figure 1 (ignoring for the moment the attached #sarcasm hashtag). Without knowing the author’s political inclination, it would be difficult to conclude with certainty whether the tweet was intended as sarcastic or not. This task is about the binary classification of tweets in English based on the category of ironic speech. As a Multilabel-Classification, a tweet can belong to multiple categories or none. This research is important because social networking sites have changed our lives in recent years. For companies, this social reality has turned into an obligation to set up communication and marketing channels through social networks. Besides, companies require feedback from consumers, through comments or messages that mark the acceptance or rejection of each of the proposals, products or services. For this competition 2 models/strategies were • used: The first model is a binary multilabel classifier using Bayesian networks. • The second model and final submit on the competition consisted of six different binary classifiers using BERT. In other words, one for each evaluable label: sarcasm, irony, satire, I2C at SemEval-2022 Task 6: Intended Sarcasm in English using Deep Learning Techniques Adrián Moreno Monterde Laura Vázquez Ramos Escuela Técnica Superior de Ingeniería Escuela Técnica Superior de Ingeniería Universidad de Huelva (Spain) Universidad de Huelva (Spain) adrian.moreno521@alu.uhu.es laura.vazquez005@alu.uhu.es Victoria Pachón Álvarez Jacinto Mata Vázquez Escuela Técnica Superior de Ingeniería Escuela Técnica Superior de Ingeniería Universidad de Huelva (Spain) Universidad de Huelva (Spain) vpachon@dti.uhu.es mata@uhu.es Figure 1: An illustrative tweet. 856 understatement, overstatement and rhetorical question. In this task …

Content truncated to 500 words