Sarcasm is a type of irony that expresses contempt, but conversely, is beneficial in enhancing the creativity of participants (Gino, 2015). Rather than using conventional means to convey sarcasm, like facial expressions and changes in intonation, writers online must be creative and utilize lexical features to exaggerate their intent for sarcasm. The Gaussian and Multinomial Naive Bayes classification models were chosen for this study because probabilities are calculated based on independent features. In this study we examined how the accuracy in sarcasm detection differs between the Gaussian Naïve Bayes and Multinomial Naïve Bayes classifiers. We were also interested in which lexical features (adjectives, adverbs, and interjections) were more prevalent in sarcastic tweets versus non-sarcastic tweets.