Analytical comparison on detection of Sarcasm using machine learning and deep learning techniques

Abstract

Sentiment Analysis is used in Natural Language processing to detect the opinion of the text/sentence put in by the user. A lot of challenges are faced while detecting the sentiment and one of them is the presence of sarcasm. Sarcasm is very difficult to detect and there could be ambiguity about the presence or absence of sarcasm. Various rule based methods have been used in the past by researchers to detect sarcasm. However, the results have not been promising. The models developed using machine learning classifiers have gained popularity over the statistical and rule based methods. Recently, deep learning techniques have been popularly used to detect the presence of sarcasm. In this paper, we have used eight machine language classifiers such as Naïve Bayes, Support Vector Machine, etc. to detect sarcasm. Deep learning techniques are also been used along with the machine learning techniques. An ensemble model has also been trained and tested on both the datasets. Bidirectional Encoder Representations from Transformers technique has given the best performance among the deep learning and machine learning techniques with an accuracy score of 92.73% and f-score of 93% on the news headlines dataset and an accuracy score of 75% and f-score of 74% on the reddit dataset.