Comparative Analysis of Deep Learning Methods in the Realm of Sentiment Analysis

Abstract

Recent advances in deep learning have suggested number of methods which can be employed in several domains. Text classification is one of the most common natural language processing tasks and have given relevant results at the level of text classification to perform sentiment analysis. This paper compares the efficacy of different algorithms used to perform sentiment analysis. The comparison offers a global vision to contribute to a relevant system that can evaluate the different types of sentiment analysis by a Corpus (restaurant reviews). In our study we have used word embedding techniques to compare the efficacy of the simple RNN., LSTM., and BERT neural networks in the context of sentiment analysis. This research indicates that the use of BERT and LSTM yields the better outcomes., although BERT requires a longer training period.