Detecting Sarcasm in Reddit Comments: A Comparative Analysis

Abstract

Sarcasm is the use of sarcastic words to mock or mockingly show disdain for something. Several people frequently use it on social media sites like Reddit and Twitter. This study investigates the effectiveness of deep learning and machine learning algorithms in detecting sarcasm using SARC dataset consisting of 1.3 million Reddit comments with almost equal amounts of sarcastic and neutral comments. We compare several well-known machine learning classification methods, including Logistic Regression, Naïve Bayes, Decision Tree Classifier, and Convolutional Neural Networks (CNN). Our results, with an accuracy of 73.2%, demonstrate that the model designed using a fusion of CNN and Long Short-Term Memory Networks (LSTM) techniques performed better than alternative classification algorithms. Our findings show how machine learning techniques will be used in the future to identify sarcasm on social networking websites.