Sentiment Analysis for Online Shopping Reviews Using Machine Learning

Everyday shoppers need reliable and insightful reviews of e-commerce websites to enhance their shopping experience. This research study explores sentiment analysis on Amazon reviews. It utilizes them as a diverse repository of customer opinions by unlocking their embedded sentiments, thereby recognizing their pivotal role in guiding potential buyers. Sentiment misinterpretations may result from many machine learning models that have trouble comprehending the context of Amazon reviews, particularly regarding subtle wordings, sarcasm, or irony. Additionally, these models can have biases that skew sentiment analysis results, mainly when working with a diverse set of Amazon review datasets. To overcome these, three machine learning models, namely, Bidirectional Encoder Representations from Transformers (BERT), Bidirectional and Auto-Regressive Transformers (BART), and Generative Pre-trained Transformers (GPT) are used in this study. During the experimental research, it was observed that BERT gave the highest accuracy of 90% when compared with BART (70%) and GPT (84%) models. BERT’s bidirectional contextual comprehension at identifying subtleties in language provides a thorough and realistic representation of the sentiments of Amazon users, which is why the model gave the highest accuracy.