Affective Dependency Graph for Sarcasm Detection

Abstract

Detecting sarcastic expressions could promote the understanding of natural language in social media. In this paper, we revisit sarcasm detection from a novel perspective, so as to account for the long-range literal sentiment inconsistencies. More concretely, we explore a novel scenario of constructing an affective graph and a dependency graph for each sentence based on the affective information retrieved from external affective commonsense knowledge and the syntactical information of the sentence. Based on it, an Affective Dependency Graph Convolutional Network (ADGCN) framework is proposed to draw long-range incongruity patterns and inconsistent expressions over the context for sarcasm detection by means with interactively modeling the affective and dependency information. Experimental results on multiple benchmark datasets show that our proposed approach outperforms the current state-of-the-art methods in sarcasm detection.