Sarcasm Detection Using Recurrent Neural Network

Sarcasm is a sharp and often ironic utterance that is meant to convey contempt or mock. In today's world, one of the challenging problems for opinion mining task is sarcasm detection. Many researchers are exploring the properties of sarcasm like semantic properties, syntactical properties, lexical feature etc, to design algorithms for sarcasm detection. We aim at using a recurrent neural network (RNN) model for sarcasm detection because it automatically extracts features required for machine learning approaches. Along with the recurrent neural network, this model also uses long short-term memory (LSTM) cells on tensorflow to capture syntactic and semantic information over Twitter tweets to detect sarcasm. Finally, we present the result of this model and a statistical overview of the dataset.