Volume : VII, Issue : IX, September - 2018

Twitter sentiment analysis using Capsule Nets and GRU

Maisnam Niranjan Singh, Y. S Kumaraswamy

Abstract :

Sentiment analysis is one of the most researched areas in computer science.Opinion forming has been part of human behavior from the beginning but to teach a machine to conduct sentiment analysis is a  hard problem solving task. The reason why it is hard for machine to do sentiment analysis stems from the fact that language could be very ambiguous . Sentiment analysis is  all about understanding the language and  classifying the language  into something negative ,positive or neutral or like and not like. In this paper we tackle the sentiment analysis problem taken data from social networking site Twitter. In this paper we use deep learning approach to teach the machine to do sentiment analysis. We choose Capsule Net along with Bidrectional GRU , a form of recurrent neural network to do the sentiment analysis and use the Glove word embedding  vectors . A case study is presented to illustrate and in depth study of the problem and we propose certain optimization steps to reach a high accuracy for the problem to be solved. Capsule Net a differs from the traditional convolution neural network  which takes into account  spatial correlation between the components whereas Convolution neural network cannot take spatial correlation, thus even though it works well there is still some disadvantage using convolution neural network.

 

Keywords :

Article: Download PDF   DOI : 10.36106/ijsr  

Cite This Article:

Maisnam Niranjan Singh, Y.S Kumaraswamy, Twitter sentiment analysis using Capsule Nets and GRU, INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH : Volume-7 | Issue-9 | September-2018

Number of Downloads : 273

References :