Using LSTM for Context Based Approach of Sarcasm Detection in Twitter
Siti Khotijah, Jimmy Tirtawangsa, Arie Ardiyanti Suryani
Abstract
In this research, we propose a sarcasm detection by taking into consideration its many varying contexts, related to the word or phrase in a tweet. To get the related context, we extract the information with paragraph2vec to simplify the process of finding the contextual meaning. The result paragraph2vec will provide the features to help classification in Long Short Term Memory (LSTM). Long short-term memory (LSTM) is an artificial recurrent neural network (RNN) architecture used in the field of deep learning. We applied a sarcasm detection method to identify sarcasm in two different languages: English and Indonesian and classification with balanced and imbalanced data. It aims to measure the reliability of the proposed approach and how effective the method is in detecting sarcasm. The result of the experiment shows that in Indonesian, balanced data has a good accuracy of 88.33 % and imbalanced data of 76.66 %, whereas in English the balanced data has an accuracy of 79% and imbalanced data of 54.5%.