Sentiment analysis of sarcasm detection in social media

Abstract

When people talk or write more especially on social media, they tend to sometimes, flip the polarity of their expressions. Oftentimes, this change of mood affects the accuracy of the sentiment analysis. Sarcasm is a sophisticated form of irony widely used in social networks and micro blogging websites and is usually a type of sentiment where person speaks the contradictory of what individual word means, expressing gloomy feelings applying positive words. However, it is hard even for humans to recognize. Hence, sarcastic expressions play a vital role in the outcome of automatic sentiment detection in sentiment analysis. To improve on the performance of automatic detection of sentiments in social media networks, there is need for sarcastic statements to be recognised and classify correctly. Sentiment Analysis is a method for identifying people’s opinion, attitude, sentiment, and emotion towards any specific target such as individuals, events, topics, product, organizations, services etc. This paper proposed an ensemble classification method having base classifiers as Random Forest, K-Nearest Neighbour and Naive Bayes for detecting sarcasm using lexical, pragmatic, linguistic incongruity and context incongruity features. The same ensemble method and Support Vector Machines (SVM) were used to identify the sarcasm types so as to diagnose the mood of the users. We report the results and present a comparative evaluation of the ensemble method and SVM classifiers for sarcasm and sarcasm type detection indicating their suitability for the task. The results obtained from the set of experiments conducted indicate the reliability of our approach.