Customer Reviews for Aspect-Based Sentiment Analysis Using Machine Learning

Abstract

Sentiment analysis of product reviews on e-commerce websites helps to identify customer preferences. Aspect-Based Sentiment Analysis (ABSA) goes a step further by pinpointing specific aspects of a product and analyzing the sentiment associated with each, offering a more granular understanding of customer attitudes. This method improves the traditional rating-based recommendation system by focusing on product aspects. A labeled dataset is necessary to train supervised machine learning models to automate ABSA. However, their availability could be improved due to the manual effort required to create such datasets. This annotated dataset contains customer reviews of the Apple iPhone 11, which have been manually labeled with predefined aspect categories and corresponding sentiments like positive, Negative, and Neutral. The accuracy of this dataset has been validated using several state-of-the-art machine learning techniques, including Naive Bayes, Support Vector Machine, Logistic Regression, Random Forest, K-Nearest Neighbour, and a Multi-Layer Perceptron (MLP) model built with the Kera’s API. These models enable the transition from a conventional rating-based approach to a more precise, aspect-driven analysis, leading to enhanced product re commendations based on customer reviews.