SARCASMET: EXTENSION OF LEXICON ALGORITHM FOR EMOJI-BASED SARCASM DETECTION FROM TWITTER DATA

Abstract

Since the industrial revolution, the original way of communicating; face-to-face communication has been used as a model to develop the various ways of communicating known to date. Transposing the principles and codes of natural face-to-face communication to today’s online communication is a major challenge for developers. In today's digital era, social media platforms like Twitter have become a hub for expressing opinions, emotions, and humor. Sarcasm, a form of verbal irony, is a prevalent means of communication on such platforms. However, detecting sarcasm in online content poses a significant challenge due to the absence of vocal intonations and facial expressions. This necessitates the development of reliable methods to automatically identify and understand sarcasm in tweets. Sarcasm makes use of positive lingual contents to convey a negative message. Different types of approaches have been developed to implement sarcasm detection on online communication platforms. However, the levels of efficiency of these approaches have been the principal worries of developers. Lexicon algorithm is used to determine the sentiment expressed by a textual content. This sentiment might be negative, neutral, or positive. It is possible to be sarcastic using only positive or neutral sentiment textual contents. Hence, lexicon algorithm can be useful but insufficient for sarcasm detection. It is necessary to extend the lexicon algorithm to come up with systems that would be proven efficient for sarcasm detection on neutral and positive sentiment textual contents. In this work, two sarcasm analysis systems both obtained from the extension of the lexicon algorithm have been proposed for that sake. The first system consists of the combination of a lexicon algorithm and a pure sarcasm analysis algorithm. The second system consists of the combination of a lexicon algorithm and a sentiment prediction algorithm. Finally, the proposed model aims to detect the sarcasm from the text and emotion icon with improved efficiency.