An Analytical Approach for Twitter Sarcasm Detection Using LSTM and RNN

In the originality of sarcasm, its intended sense of the tweet written by the author is opposite to the literal meaning. Because of its intricate nature, it is much difficult, similarly without proper context human is not able to identify sarcasm in any verbal or textual sentence. In the recent time peoples on social media platform like Twitter, Facebook, WhatsApp succeed in recognize sarcasm despite interacting with strangers across the world. Sarcasm detection in Hindi Language text is the challenging task in the NLP due to the richness of morphology and low availability of resources. Present paper will analysis and calculate the performance of recurrent and convolution neural network model for detecting sarcasm in tweets by analysing the current state art of the performance and qualitative analysis of network functionality. In the proposed work sarcasm detection is done by binary classifier. In the present paper evaluation over 3 models 2 Recurrent Neural Networks (RNNs), 1 with LSTM-cells and 1 with GRU cells and also a CNN. For the analysis of network functionality, classification result will compare with resultant BoW models, jumbling the word content in tweets and understand the network decision. Experimental results shows that network prediction precisely based on the occurrence of word’s in the tweets and the performance based on F1 score reaches up to 85% using RNN and CNN models. Suggested model performs in a well way in comparison of Bag of Words BoW model and ensures the feasibility of neural network in sarcasm detection in tweets.