GetSmartMSEC at SemEval-2022 Task 6: Sarcasm Detection using Contextual Word Embedding with Gaussian model for Irony Type Identification

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 827 - 833 July 14-15, 2022 ©2022 Association for Computational Linguistics GetSmartMSEC at SemEval-2022 Task 6: Sarcasm Detection using Contextual Word Embedding with Gaussian Model for Irony Type Identification Diksha Krishnan and C.Jerin Mahibha Meenakshi Sundararajan Engineering College, Chennai dikshakrish5@gmail.com jerinmahibha@gmail.com Durairaj Thenmozhi Sri Sivasubramaniya Nadar College of Engineering , Chennai theni_d@ssn.edu.in Abstract Sarcasm refers to the use of words that have different literal and intended meanings. It rep- resents the usage of words that are opposite of what is literally said, especially in order to in- sult, mock, criticise or irritate someone. These types of statements may be funny or amusing to others but may hurt or annoy the person to- wards whom it is intended. Identification of sarcastic phrases from social media posts finds its application in different domains like sen- timent analysis, opinion mining, author pro- filing and harassment detection. We have pro- posed a model for the shared task iSarcasmEval - Intended Sarcasm Detection in English and Arabic by SemEval-2022 considering the lan- guage English. The Subtask A and Subtask C were implemented using a Convolutional Neu- ral Network based classifier which makes use of ELMo embeddings. The Subtask B was implemented using Gaussian Naive Bayes clas- sifier by extracting TF-IDF vectors. The pro- posed models resulted in macro-F1 scores of 0.2012, 0.0387 and 0.2794 for sarcastic texts in Subtasks A, B and C respectively. 1 Introduction In the Internet era, specifying user comments, views and opinions through social media has be- come very common and these may not be speci- fied directly and can include indirect phrases with implicit meanings (Abu Farha and Magdy, 2020). The posts may also represent undesirable charac- teristics using positive words and they may not be formal in nature. It is common to use abbreviations, uncommon, ambiguous and multilingual words in social media posts and no predefined structures are explicitly defined for sarcastic messages (Sarsam et al., 2020). So the identification of sarcastic posts cannot be considered as a direct process. Exhaus- tive understanding of the contextual meaning of the posts is considered as an important factor in identi- fying sarcastic posts. Sarcasm detection is consid- ered as a challenging task associated with sentiment analysis. Sarcasm detection also plays an important role in analysing the voice of the customer based on which major decisions will be taken. Sarcasm requires some shared knowledge be- tween speaker and audience and it is considered as a profoundly contextual phenomenon. Using sarcastic phrases in sentences is a common way of ironic or satirical speech for the common man. That being said, social media platforms, such as Twitter, Facebook and YouTube contain millions of tweets and comments that include sarcastic phrases, thus making it a field of study under the domain of NLP. For example, a user can change a supposedly negative comment using positive words as in the sentence, “It is awesome to go to bed at 3 am #not”. In such cases …

Content truncated to 500 words