An Optimized Approach for Sarcasm Detection Using Machine Learning Classifier

Sarcasm is a sardonic or bitter statement used to express contempt. When sarcasm is used, the user's original intent for their tweet is often quite different from its literal interpretation. It is significantly hard due to its complicated nature, much as a person cannot recognize sarcasm in any verbal or written sentence without the correct context. Despite communicating with strangers across the world, users of social media platforms like Twitter, Facebook, and WhatsApp are currently successful at identifying sarcasm. The research objective of the present paper is to detect sarcasm in social media tweets using machine learning classifier models. To identify sarcasm, the current study employed logistic regression, a naive Bayes classifier, a linear support vector machine, a decision tree, and an ensemble classifier. These models are put into practice in Python and tested over different datasets, both sarcastic and non-sarcastic. The effectiveness of the models is evaluated in terms of several measures, including precision, accuracy, recall, and the F-measure. This paper uses logistic regression, a naive Bayes classifier, a linear support vector machine, a decision tree, and ensemble classifier models to detect sarcasm. By using this classifier model and linguistic context, our model achieves a significant accuracy around 0.75 in identifying sarcastic tweets. The proposed work contributes to the real-time application of natural language processing to detect sarcasm in social media tweets.