Comprehensive Sentiment Analysis for Mining Consumer Opinion in Online Product Rating

This study conducts a comprehensive analysis contrasting two distinct sentiment analysis models: VADER and Roberta, a sophisticated pre-trained deep learning model integrated via the Transformers library pipeline. Emphasizing their performance in predicting sentiment from product ratings and reviews, this research aims to provide a nuanced understanding of their respective methodologies. Initially, an extensive review of VADER's rule-based approach and Roberta's advanced pre-trained architecture is presented, highlighting unique strengths and limitations inherent in each model. Subsequently, both models undergo rigorous training and evaluation using a rich dataset of product ratings and reviews, consisting of 500 readings carefully selected from the larger dataset of 568,454 readings. Rigorous assessment metrics, including accuracy, precision, and recall, are employed to gauge their proficiency in sentiment classification. The study's outcomes shed light on the applicability and efficiency of NLTK-anchored VADER and Roberta within the e-commerce sphere, contributing valuable insights into leveraging distinct models for extracting actionable customer insights. Specifically, VADER achieved an accuracy of 82%, precision of 78%, and recall of 85%, while Roberta demonstrated higher performance with an accuracy of 88%, precision of 85%, and recall of 90%. By exploring product rating sentiment analysis, this research facilitates informed, data-driven decisions for businesses, ultimately enhancing customer satisfaction and refining product development strategies for sustainable growth. The findings underscore the importance of leveraging advanced deep learning techniques for robust sentiment analysis in modern e-commerce ecosystems.