Sarcasm Detection Using Lexical and Contextual Features of Deep Learning Architecture

Abstract

ADEYEMO, Adesesan Barnabas
ADEWUYI, Joseph Oluwaseyi
sesanadeyemo@gmail.com
adewuyiseyi@gmail.com
Department of Computer Science, University of Ibadan, Ibadan, Nigeria.
Abstract
In contemporary time, social media sites such as Facebook, LinkedIn, Twitter, etc. have expanded and received vast
admiration and significance. These sites have change into large environments where users express their ideas, views and
opinions naturally. Companies and organizations leverage this unique environment to tap into people’s opinion on their
products or services and to make available instantaneous customer assistance. This research seeks to avoid the use of
grammatical words as the only features for sarcasm detection but also the contextual features, which are theories
explaining when, how and why sarcasm is expressed. The contextual features consider the user’s current and previous
posts to detect or classify if a post is sarcastic. A deep neural network architecture model was employed to carry out this
task, which is a bidirectional long short-term memory with conditional random fields (Bi-LSTM-CRF), two phases were
employed to classify if a reply or comment to a tweet is sarcastic or non-sarcastic. In the first phase, classification was
carried out separately using the comment and the reply alone. In the second phase, the classification considers both the
reply and the context of the reply with the original tweet. For these two phases, experiment was carried out using the Bi-
directional Long-Short Term Memory (Bi-LSTM). The inclusion of Conditional Random Field (CRF), which is a
probabilistic model for structured prediction helped to predict using the output of both forward and backward propagation
of the LSTM. The performance of the models was evaluated using the following metrics: Accuracy, Precision, Recall, F-
measure. The model has 0.9211 Accuracy, 0.92134232 precision, 0.9122 recall and 0.9131832 F-measure.