Sentiment Analysis of Customer Reviews from E-Commerce Website

In the rapidly evolving e-commerce landscape, understanding customer sentiment is crucial for enhancing products and services. This work involves a sentiment analysis system that not only categorizes sentiments in real-time but also leverages a models combining lexicon-based and transformerbased techniques to improve accuracy and handle contextual nuances more effectively. This study uses the amazon fine food reviews dataset for the analysis with over 4500 reviews. We compare traditional machine-learning approaches, including Naïve Bayes with 80.91 % accuracy, logistic regression with 83.58 % accuracy, VADER with 82.8 % accuracy and deep learning-based model Roberta with 87.0 % accuracy. By integrating these advancements, the system would provide businesses with timely, accurate insights, enabling them to quickly adapt to customer feedback. The data-driven insights generated will support strategic decision-making in product development, marketing, and customer service, offering businesses a competitive edge and fostering long-term success in the digital marketplace through continuous, real-time improvement based on customer feedback.