Study on Text Classification using Capsule Networks

Abstract

Deep Neural Networks are being used in different domains, from Image Classification to Natural language Processing. Research has been done on Artificial Neural Networks and different types of Neural Networks such as CNN and RNN have been studied and developed. They have been applied in different applications. In 2017, a new concept has been introduced in the Neural Networks Architecture by Geoffrey Hinton – Capsule Networks.Capsule Networks bring an improvement in the old neural network architecture and it has worked better than the Convolutional Neural Networks (CNN). There are certain disadvantages of using the convolutional neural networks, few areas where CNN lacks, Capsule Networks have overcome those limitations and now it is better neural network architecture for developing models to solve the problems in different domains. Capsule Networks are primarily used for Image Classification. They can be applied in the areas of Natural Language Processing and Recommender Systems to utilize the textual information in a more efficient manner. Neural Networks can be trained to learn numeric representations of various words and phrases using models such as Word2Vec and Glove and can be applied to classify the data in different categories. This paper presents a review on text classification using the newly introduced capsule networks.