BERT Model Adoption for Sarcasm Detection on Twitter Data

Abstract

Sarcasm is a term used to criticize someone's feelings. Sometimes, humans are not able to identify sarcastic comments, and they typically express the reverse of what they mean when they make snarky remarks. Therefore, the detection of sarcasm within a text automatically is a difficult task. Its significance in enhancing sentiment analysis has also made it an important study field. In previous studies, different approaches to deep learning (DL) and machine learning (ML) have been explored. However, previous approaches mainly depend on the lexical and linguistic aspects. Therefore, these techniques could not perform well in the context of sentiment accuracy. In this research, an efficient approach for detecting sarcasm is proposed. A Bidirectional Encoder Representation from a Transformer (BERT) is proposed to improve the sentiment accuracy in this research. This research also aims to compare the two models of deep learning, the BERT and LSTM (Long Short-Term Memory) models. This comparative analysis aims to provide a detailed overview of the pros and cons of each approach for the detection of sarcasm. The primary aim of this study is to examine the different existing ML and DL approaches for the identification of sarcasm. Apart from this, the comparison of BERT and LSTM contributes to the ongoing debate about whether models work best for sarcasm detection in social media. In this study, sentiment analysis's accuracy is improved by making better decisions, especially when it concerns Twitter interactions.