Decoding sarcasm: unveiling nuances in newspaper headlines

Abstract

This study navigates the intricate landscape of sarcasm detection within the condensed confines of newspaper titles, addressing the nuanced challenge of decoding layered meanings. Leveraging natural language processing (NLP) techniques, we explore the efficacy of various machine learning models—linear regression, support vector machines (SVM), random forest, na¨ıve Bayes multinomial, and gaussian na¨ıve Bayes—tailored for sarcasm detection. Our investigation aims to provide insights into sarcasm within the succinct framework of newspaper titles, offering a comparative analysis of the selected models. We highlight the varied strengths and weaknesses of these models. Random forest exhibits superior performance, achieving a remarkable 94% accuracy in accurately identifying sarcasm in text. It is closely trailed by SVM with 90% accuracy and logistic regression with 83% accuracy.