Sarcasm recognition is another aspect of natural language processing (NLP) used for sentiment analysis. It uses mathematical inference to show and categorize a word or phrase’s polarity for sardonic. In literature, sarcasm detection has been applied to various datasets such as the Reddit corpus and Twitter datasets, which mostly contain noisy language and labels. This work aims to identify sarcasm in the following datasets: news headlines, generic, rhetorical questions, hyperbole, English and Arabic text to overcome these issues. This study compares various models intended to distinguish between sardonic and non-sardonic information to understand this intricate problem efficiently. Moreover, standard datasets are trained using deep learning techniques, including transformers like GPT and BERT. The transformer-based models (GPT and BERT) outperform other models, with F1-scores of 87% and 85%, respectively. The model with the lowest performance is the logistic regression model, having a F1-score of 73%.