Sarcasm Detection with and without #Sarcasm: Data Science Approach

Abstract

Natural languages usually contain context, which is difficult for a machine to understand. Sentiment analysis is a contextual mining technique often used in NLP to identify, understand and extract subjective information in texts, such as people’s comments, feedback, reviews, and opinions. Sentiment analysis is a useful tool for finding the polarity of a sentence. Sarcasm detection is one of the complex areas of sentiment analysis. Sarcasm flips the polarity of the sentence identified by sentiment analysis. Thus, sentiment analysis results may get biased if people use sarcasm in their text. Hence, to understand the sentence's real meaning, we proposed a system of sarcasm detection on tweets using an ensemble approach. We performed sarcasm detection with and without #sarcasm. After training a model and observing earlier studies, We found that the presence of #sarcasm gives a better result. Therefore the author tried implementing a model where #sarcasm is removed from the tweets, and the model is trained. After comparing both models' presence and absence of hashtags, it is found that the lack of the hashtag model works well, which can be used on any plain text without any clue of sarcasm.https://dorl.net/dor/20.1001.1.20088302.2022.20.4.1.0

Keywords

• sarcasm detection, hashtag, machine learning, deeplearning