SENTIMENT ANALYSIS USING NOVEL DEEP LEARNING METHODS

Abstract

In the current digital era, the humongous amount of data being generated has been impacting public lives in one or the other ways. Sentiment analysis, also known as opinion mining, is related to contextual mining of texts which helps in identification and extraction of subjective information from the source material. Sentiment analysis is being used for brand monitoring and reputation management across different market segments. It helps to understand how the public perceive a particular brand, product or service that is highly useful for different tech companies, marketing agencies, media organizations, fashion brands etc. In today’s scenario we have been suffering with data overload which makes it impossible to analyze public sentiments without any sort of error or bias. Sentiment analysis provides better insights into the public reviews as it can be automated which ultimately helps in decision making. There are various deep learning and machine learning methods and models as well as natural language processing tools which help in examining and analyzing public opinions with low time complexity. However, deep learning methods have become highly popular in recent times as these models provide high efficiency and accuracy. In this review paper we have provided a complete overview of the common deep learning frameworks being employed for sentiment classification and analysis. This paper discusses various learning models, evaluation, text representations and other metrics in deep learning architectures. The key findings of different authors have been discussed in detail. This paper will help other researchers in understanding the deep learning techniques being used for sentiment analysis.

Keywords

• sentiment analysis
• machine learning
• deep learning
• natural language processing
• deep neural networks (dnn)
• convolutional neural networks
• recurrent neural networks (rnn)