Aspect-Based Sentiment Analysis for Online Products

Abstract

A number of resources are now available for gauging public opinion about businesses, products, services, brands, and events because of the advent of e-commerce. Customers may find it difficult to distinguish between legitimate assessments and biased or false reviews, which leads to worse goods or services. Aspect-based sentiment analysis is utilized in this study to develop a natural language processing model to examine customer reviews and comments to ascertain the sentiment expressed towards certain qualities or components of a good or service. A few procedures involved in this undertaking include aspect extraction, sentiment classification, training, feature extraction, and dataset preparation. The T5 model will be used to train the model using a dataset of customer reviews, and it will be evaluated using a variety of performance measures, including accuracy (84.62%), precision (84.58%), recall (84.62%), and F1-score (83.39%). The project's findings will help customers locate services that meet their needs, and they will also assist businesses in learning more about the advantages and disadvantages of their goods and services.