Very Deep Convolutional Neural Network Based Sarcasm Sentiment Detection and Classification Model On Twitter

Authors

  • K. Kavitha , Research scholar, Department of Computer Science & Engg., Acharya Nagarjuna

Abstract

At present times, in the general area of sentiment analysis, sarcasm plays a task as an interfering
factor which will flip the polarity of a message andusers are frequently posting sarcastic messages on
social networking sites such as Twitter, Facebook, WhatsApp, etc. The process of identifying the
sarcastic messages has become an essential process since they have great impact on diverse fields.
For effective sarcasm detection, an automatic sarcasm recognition model has been designed to
identify sarcastic posts on social networks for recognizing the original sentiment of a given text in
the existence of sarcasm. In this paper, a new feature extraction dependent classifier technique has
been presented to detect the sarcasm in Twitter. A set of features extracted are sentiment,
punctuation, syntactic, semantic features, and pattern features. Then, the extracted features undergo
classification by the use of Very Deep Convolutional Neural Networks (VDCNN) which classifies
the tweets into sarcastic and non-sarcastic ones. VDCNN works straightaway on the character level
and utilizes less number of convolution and pooling operations. The simulation analysis takes place
to ensure the goodness of the VDCNN model and the results exhibits the supremacy nature of the
presented model over the compared methods.

Published

2020-12-01

Issue

Section

Articles