METHODOLOGIES FOR SARCASM DETECTION ON ONLINE SOCIAL MEDIA

Abstract

Sarcasm Detection is procedure to identify the phrases that express a meaning contrary to what it really wants to express. The metaphorical nature of sarcasm presents a significant difficulty for sentiment analysis systems based on emotion detection. Detecting sarcasm on online forums is a very specific topic in natural language processing (NLP), a sort of sentiment analysis that focuses on recognising sarcasm rather than discovering a perception over the whole domain. Sarcasm identification and sentiment analysis vary by a hair's breadth. In NLP, sarcasm recognition is a rather specific research topic. As a consequence, the objective is to figure out if a script or phrase is sarcastic or not. In our previous survey we have discussed and compared the approaches to sarcasm detection. This research paper is focused on how elegantly to analyse the large datasets without losing the efficiency of the Deep Learning models. In recent years there is significant increase in the social media users that is why identifying sarcasm is of great concern. Sarcastic remarks in the form of tweets typically include positive phrases that symbolise bad or unpleasant attributes, therefore recognizing sarcasm on social media has gotten a lot of attention recently. In this paper we implement Knowledge distillation approach to detect sarcasm in news headline. Proposed methodology reduces the complexity of the deep learning neural network by retaining the accuracy.