Deep Models for Converting Sarcastic Utterances into their Non Sarcastic Interpretation
Abhijeet Dubey, Aditya Joshi, Pushpak Bhattacharyya · 2019
Sarcasm is a form of speech in which the the implied sentiment is the opposite of literal meaning. In this paper, we present the task of sarcasm interpretation, defined as converting a sarcastic utterance into its non-sarcastic (literal) interpretation. We present three approaches for the task: (a) a rule-based approach that considers sarcasm as a form of dropped negation and associate negation words with verbs present in the sarcastic utterance, (b) statistical machine translation-based (SMT) approach that address the sarcasm interpretation task as monolingual machine translation and (c) three deep learning-based (DL) architectures, Encoder-Decoder Network, Attention Network and Pointer Generator Network. We also discuss the scope of future work to further enhance the proposed models for sarcasm interpretation.