A Hybrid Framework for Sarcasm Detection Using ‎CB Technique

In the last two decades, people have been posting about their choices, likes, dislikes, their views, and their opinions on websites. People tend ‎to post text feedback, audio feedback, and video feedback on websites. Natural language processing techniques such as emotion detection, ‎sentiment analysis, etc., can be used to judge the sentiments of the people. Businesses can use these techniques to judge their products and ‎services and, in turn, decide the respective future actions to be taken based on the opinions. In sentiment analysis, the feedback can be judged ‎as positive, negative, or neutral. Some people express their views sarcastically. By being sarcastic, the opinion is inverted in terms of polarity. ‎Neural network techniques have been used recently for sarcasm detection. This study is focused on detecting sarcasm from text reviews ‎using the CB technique of Convolution Neural Networks (CNN) and Bi-directional Encoder Representations from Transformers (BERT). ‎This technique improves the performance compared to the individual deep learning techniques by having an F-score of 95%‎.