A Statistical Approach for Sarcasm Detection Using Twitter Data
Rahul Gupta, Jitendra Kumar, Harsh Agrawal, Kunal
Abstract
Sarcasm is an alternate sort of assessment that comprises of words which by and large methods something contrary to what you truly need to state and can be utilized in various circumstances, to affront, to show disturbance, or be clever. Here and there to pass on understood data, sarcasm is utilized inside the message as an individual conveys. Sarcasm can be used for different capacities, similar to analysis or joke. In any case, sarcasm is extremely debilitating in any event, for people to recognize. Therefore, the detection of sarcasm leads to a better understanding of sentiment analysis of the user, from the data collected from websites such as Twitter and Facebook. The advantages of identifying sarcasm for opinion mining has prompted enthusiasm for automated sarcasm recognition as an exploration issue. In Automated sarcasm identification, the program tries to identify whether a text is sarcastic or neutral. The research paper is divided into two phases. In the first phase, it extracts features related to sentiments and punctuation, and then the chi-square test is used to shortlist the most useful features. In the second phase, 200 top tf-idf features are extracted and combined with sentiment related and punctuation related features to find sarcastic content within the tweet.In the first approach the best accuracy is achieved by using the Support Vector Machine algorithm which is equal is equal to 74.59% and in the second approach the maximum accuracy is achieved by the voting classifier and it increases the accuracy to 83.53%.