Sentimental Analysis on Amazon’s Alexa Reviews using ML and DL approach: A Comparative study

Abstract

Sentimental analysis is a processes of mining textual data like reviews, statements and make prediction about emotions behind it. Businesses today use sentimental analysis to monitor their product popularity, customer’s perception about product through feedback. Amazon’s alexa is one of the virtual assistants increasingly adopted by people. Despite of its popularity, there are mixed reactions among people regarding alexa and its variations. In order to get good insight of alexa and its variants, it is important to go through the written reviews and proper analysis of these written review should be done, but doing it manually will be a time consuming and cumbersome task. Thus, an automatic sentimental analyzer is required which will benefit the buyer as well as a manufacturer. This work attempts to develop a sentimental analyzer which analyzes 5000 Alexa reviews using naïve bayes algorithm, Random forest, Long Short-Term Memory Networks (LSTM) and multilayer perceptron (MLP).It further evaluate performance of each algorithm with respect to various parameters like precision, recall, F1 score and ROC and a brief comparative study is presented.