Sentiment Analysis of Online Product Reviews using ML, DL and XAI Techniques: A Systematic Review

Abstract

With the advent of Web 2.0 along with affordable internet data rates as well as smartphones, many people are doing online shopping and writing reviews on shopping sites. These reviews have become increasingly vital for both consumers and sellers. Online customers and their reviews are increasing abruptly. Since customer reviews are unstructured in nature, it is difficult to unsheathe the sentiment from them. The term sentiment analysis (SA) refers to a technique that analyzes and detects the emotions, perspectives, attitudes, and sentiment of individuals hidden in review text. This paper examines and documents previous work in SA related to online product reviews using Machine Learning (ML), Deep Learning (DL), and Explainable Artificial Intelligence (XAI). Study identifies various sentiment analysis levels, approaches, datasets and feature extraction techniques applied in past work. The findings of this review revealed that the most widely used SA approaches for this domain are the Support Vector Machines (SVM) from Machine Learning and the Convolutional Neural Networks (CNN) from Deep Learning. Research in DL approaches can be extended using Hybrid DL models with novel word embedding methods. XAI methods can be used to make opaque DL models more interpretable in turn trustworthy.