Deep Learning Based Sentiment Analysis on Product Reviews on Twitter

Abstract

Sentiment analysis is the process of extracting an opinion about a particular subject from text documents. The immense quantity of text documents contain opinions or reviews towards a particular entity. The identification of sentiment can be useful for individual decision makers, business organizations and governments. Sentiment analysis is an important research direction. Deep learning is a recent research direction in machine learning, which builds learning models based on multiple layers of representations and features of data. Deep learning based frameworks can be employed in a wide range of applications, including natural language processing tasks, with encouraging prediction results. In this paper, we present a deep learning based scheme for sentiment analysis on Twitter messages. In the presented scheme, three-word embeddings based schemes (namely, GloVe, fastText and word2vec) and convolutional neural network (CNN) have been utilized. In the empirical analysis, different subsets of Twitter messages, ranging from 5000 to 50.000 are taken into consideration. The prediction results obtained by deep-learning based schemes have been compared to conventional classifiers (such as, Naïve Bayes and support vector machines).