Sarcasm Detection Using Deep Learning Approaches: A Review

Abstract

Emotions are something that makes one realize how other people are feeling but sarcasm needs to be understood by putting in some extra effort. Sarcasm, a verbal irony, is a practice of using words or sentences that are different from their literal meaning. Researchers are still making effort in developing an algorithm that can identify sarcasm completely. Since sometimes humans also take time to understand sarcasm, making a machine learn to recognize is also not a simple task. The need for Deep Learning (DL) is rapidly growing for detection and classification operations. Different research works focused on Sarcasm detection using various methodologies but the issue with existing research work is their performance and accuracy. Our survey provides several helpful examples, the most notable of which is a table that lists prior studies according to several criteria, including the kinds of methodologies with accuracy, and datasets employed. This paper also throws light on multimodal detection, sarcasm detection from typographic images (memes), feature set analysis, and different phases of a model with various issues and milestones in sarcasm detection.