A Combinational Approach for Sarcasm Detection in Twitter

Sentiment analysis, also known as opinion mining is the qualitative method that uses natural language processing, text analysis and computational linguistics. The goal of sentiment analysis is to determine if a specific passage in the text shows positive, negative or neutral sentiment towards the subject. The main objective is to detect and analyze the emotional reactions of the speaker or writer based on his attitude. These emotions may change the polarity of the text and make it sarcastic. Due to these difficulties and the inherently tricky nature of sarcasm, it is generally ignored during social network analysis. As a result, the outcome of such analysis is affected adversely. Thus, sarcasm detection poses to be one of the most critical problems which we need to overcome while trying to yield high accuracy insights from abundantly available data. In this paper, we propose a combination of existing techniques to detect sarcasm on Twitter. It involves an effective preprocessing approach which are combined with four sets of features that can handle different types of sarcasm. We insist that the usage of semi-supervised learning method to train the sarcasm detection model which will increase the accuracy by approximately 3 % than the state-of-art approaches. Hence, the proposed work may provide precision, recall and F-score around 90 %, 96.5 %, 92.0 % respectively.