Sarcasm Detection for Marathi and the role of emoticons

Sarcasm detection has gained a lot of attention in the the natural language processing (NLP) regime, largely due to the complexity involved in identifying and modeling the actual sarcastic intent behind a sentence or text. This problem has also been actively studied in the context of the posts on social media platforms such as Twitter. Several works in this area are available for English and a number of foreign languages. As far as the Indian languages are concerned, some research has also been conducted toward languages such as Hindi and Tamil. However, despite the fact that Marathi is the third most popular language in India, sarcasm recognition in Marathi remains unexplored. Most existing sarcasm detection algorithms focus on textual information, while ignoring rich semantic information expressed by the user in the form of emoticons (emojis). In real-world scenarios, emojis are often perceived as emotion signals, which strongly indicate the intent behind the text, and thus can potentially be used to improve sarcasm detection. In this paper, the problem of sarcasm detection for Marathi language is explored. Also, the significance and effectiveness of using emojis as a strong feature for sarcasm detection using machine learning algorithms is demonstrated.