A comparative study on various pre-processing techniques and deep learning algorithms for text classification
by P. Bhuvaneshwari; A. Nagaraja Rao
International Journal of Cloud Computing (IJCC), Vol. 11, No. 1, 2022

Abstract: Pre-processing is the primary technique employed in sentiment analysis and selecting the suitable techniques for the corresponding application can increase the classifier accuracy. It reduces the complexity innate in the raw data which makes the classifier to learn faster and precisely. Despite its importance, the pre-processing in polarity deduction has not attained much attention in the deep learning literature. So, in this paper, 13 popularly used pre-processing techniques are evaluated on three different domain online user review datasets. For evaluating the impact of each pre-processing technique, four deep neural networks are utilised, and they are auto-encoder, convolution neural network (CNN), long short-term memory (LSTM), and bidirectional LSTM (Bi-LSTM). The purpose of this paper is to identify the appropriate pre-processing techniques and the best classifier which achieves higher accuracy. Experimental results of this study show that using appropriate pre-processing techniques can significantly improve the classifier accuracy. Also, it is noted that Bi-LSTM model achieves higher accuracy rate than the remaining neural networks.

Online publication date: Thu, 24-Feb-2022

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Cloud Computing (IJCC):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com