Sentiment Analysis-Based Categorized Opinions Expressed in Feedback Forums Using Deep Learning Technique and Message Queue Architecture

Introduction

Sentiment analysis technology is widely used by the industry and institutions. The aim of such a system is to find the response of the customers toward the product or service of the company. The sentiment is widely used by the industry how their product or service is received by the customers as it is important for the company to remain competitive in the market and to run profitably. The significance of this system is that instead of employing large numbers to get the sentiment of the product which can be a very time consuming process. The company can automate it and save a lot of resources and time which it can use to improve and create new products and services. In this regard, some research has been done by various researchers using sentiment analysis (Fang & Zhan, 2015; Pankaj et al., 2019). The sentiment is to get the reaction of the user i.e. how it is received by the user. The comments by the customers are classified into 5 categories on the level of how they are received by them. These categories reflect the mood of the customer toward the product, 1 is the lowest rating and 5 is the highest rating. In this study, it was also tried to know how this is generally received by the public for which the average of the rating is taken to get the general idea and what step to be taken so that its rating can be improved. But our target is not just to create a sentiment analyzer but to create a one which can be easily loaded on small devices like microprocessors which are the basis of many cloud platforms and also have a very fast response time (almost real time analysis).

So in this work, an attempt was done to create a system which can give us some good performance and also faster computation time. There is a trade-off between the performance of the model and the computation time. It was tried to achieve the optimal position where we can have the better of both sides as most companies are using cloud for their system nowadays and can’t provide higher computational time to sentiment analysis as it will cost them more so the model should have lower number of computation and high enough performance. In this direction, it was tried to create a small size model with less number of parameters and have high enough accuracy so that companies can make accurate predictions for their product with less amount of error.

Almost every business, company and institution requires a sentiment analysis method to know how well their product or service is received by the user. It is very important for companies to know what their users want, to remain relevant in the market and earn the profit. It is an integral part of the company to know these details so they can adjust their product and business strategy according to it. Since in the present age when every company has millions of users of their product it is not possible for them to know the sentiment of the users by manually reading their replies and sentiments. It will be too much labour intensive and very time consuming and every time wasted by the company will cost the company a lot of money. So a need arises to automate the process of finding the sentiment of the product so that the time can be reduced from days to minutes, this extra saved resource can be used to analyze the product and user requirement in order to improve the product for higher benefit out of it and save the company money.

So for the automation of this process to find the sentiment of the product, requires creating hard code the problem which in turn needs a lot of the conditions and code a lot of the grammar and the code will not be as flexible on the data as it is essentially to be. Therefore, a need arises to employ various soft computing techniques to find the sentiment for our product. For this process many computer scientists and researchers have come up with various models of machine learning and deep learning techniques for developing automated system sentiment analysis. They have used various techniques to solve this problem. This type of task in which some computation is performed on the language related data like text or voice come under the techniques of natural language processing (NLP) in soft computing techniques. In further sections, literature review, methodology used in work, result and discussions, and conclusion and future scope have been discussed.