Comparative Study on Sentiment Analytics Using Deep Learning Techniques

Abstract

Sentiment analytics is the process of applying natural language processing and methods for text-based information to define and extract subjective knowledge of the text. Natural language processing and text classifications can deal with limited corpus data and more attention has been gained by semantic texts and word embedding methods. Deep learning is a powerful method that learns different layers of representations or qualities of information and produces state-of-the-art prediction results. In different applications of sentiment analytics, deep learning methods are used at the sentence, document, and aspect levels. This review paper is based on the main difficulties in the sentiment assessment stage that significantly affect sentiment score, pooling, and polarity detection. The most popular deep learning methods are a Convolution Neural Network, Recurrent Neural Network and Long Short Term Memory. Finally, a comparative study is made with a vast literature survey using deep learning models.