Toward Sarcasm Detection in Reviews—A Dual Parametric Approach with Emojis and Ratings

Detection of sarcasm is very crucial in today’s world where social media become a major platform of expressing emotions. Sarcastic statements are the statements where the sentiment polarity and contextual meaning are completely contrary. It affects the efficiency and accuracy of present sentiment analysis systems (SAS). Most of the currently available sarcasm detection models such as vector space models, CNN, RNN, etc. consider the raw review text in order to determine the sentiment which ignores the presence of negation, lexical ambiguity, and irony created by general facts. Due to this, detecting sarcasm may not be very accurate. To improve the accuracy of sarcasm detection and to better understand the context, the model proposed integrates the ratings, reviews, and emojis. The performance of the proposed system is evaluated using annotator-agreement methods with the metrics such as F1 score, Precision, Recall, and Accuracy. The performance shows that integrating more features enhances the accuracy by a considerable margin as compared to previously defined methodologies.