A Survey on Automated Sarcasm Detection on Twitter
This is an incremental survey that summarizes current techniques for a domain-specific problem in natural language processing, aimed at researchers and practitioners working on social media analysis.
The paper surveys existing methods for automated sarcasm detection on Twitter, addressing the problem of miscommunication due to missing context, and notes a trend towards deep learning approaches like transformers.
Automatic sarcasm detection is a growing field in computer science. Short text messages are increasingly used for communication, especially over social media platforms such as Twitter. Due to insufficient or missing context, unidentified sarcasm in these messages can invert the meaning of a statement, leading to confusion and communication failures. This paper covers a variety of current methods used for sarcasm detection, including detection by context, posting history and machine learning models. Additionally, a shift towards deep learning methods is observable, likely due to the benefit of using a model with induced instead of discrete features combined with the innovation of transformers.