Twitter Sarcasm Detection Exploiting a Context-Based Model

Automatically detecting sarcasm in twitter is a challenging task because sarcasm transforms the polarity of an apparently positive or negative utterance into its opposite. Previous work focus on feature modeling of the single tweet, which limit the performance of the task. These methods did not leverage contextual information regarding the author or the tweet to improve the performance of sarcasm detection. However, tweets are filtered through streams of posts, so that a wider context, e.g. a conversation or topic, is always available. In this paper, we compared sarcastic utterances in twitter to utterances that express positive or negative attitudes without sarcasm. The sarcasm detection problem is modeled as a sequential classification task over a tweet and his contextual information. A Markovian formulation of the Support Vector Machine discriminative model as embodied by the $$SVM^{hmm}$$ algorithm has been employed to assign the category label to entire sequence. Experimental results show that sequential classification effectively embodied evidence about the context information and is able to reach a relative increment in detection performance.