A pragmatic and intelligent model for sarcasm detection in social media text

Abstract

The world has now become an ecumenical village because of the Internet. Online platforms like e-commerce sites, search engines, social media have convoluted with the general routine of daily life. Social sites such as Twitter and Facebook have a user population larger than most of the countries, due to which communication is now largely shifted to text-based communication from verbal communication. This research investigates a common yet crucial problem of sarcasm detection in text-based communication. To prevent this problem a novel model has been proposed based on Google BERT (Bidirectional Encoder Representations from Transformers) that can handle volume, velocity and veracity of data. The performance of the model is compared with other classical and contemporary approaches such as Support Vector Machine, Logistic Regression, Long Short Term Memory and Convolutional Neural Network, BiLSTM and attention-based models which have been reported to be used for such tasks. The proposed model establishes its competence by evaluation on different parameters such as precision, recall, F1 score and accuracy. The model is built with the hope that it may help not only the government but also the general public to build a safer and technologically advanced society.