Numerical Irony Identification: A Pattern-Based Approach

Abstract

Irony is a type of metaphorical language where the literal meaning of words can't hold, rather the contrary interpretation is planned in a content. This purposeful uncertainty makes irony recognition a significant job of opinion mining. In general irony discovery is viewed as a binary classification issue wherein both rule-based and deep learning models have been effectively worked to foresee wry remarks. These current strategies will in general focus on recognizing irony in a content. However a specific type of irony communicated through numbers stays neglected. In this manner, perceiving numerically ironic statements can be extremely valuable to improve the performance of opinion mining of information gathered from microblogging sites or interpersonal organizations. In this paper, we proposed a pattern-based methodology for recognizing irony communicated through numerical qualities in a tweet. We dissect the difficulties of the issue and present a Machine Learning way to deal with irony in numerical bits of text. Four sets of highlights that cover the various sorts of irony are characterized and used to classify the tweets as ironic and non-ironic. We likewise study the significance of every one of the proposed sets of highlights and assess its additional incentive to the classification. The exploratory outcomes show that our model can clearly beat other cutting-edge strategies and further, we underline the significance of pattern-based highlights for the recognition of ironic statementsexpressed through numbers.