Automated Sentiment Classification of Amazon Product Reviews using LSTM and Bidirectional LSTM
R Prasanna Kumar, Bharathi Mohan G, R Elakkiya, Charan Kumar M, M Rithani
Abstract
Sentiment analysis is crucial for understanding opinions and sentiments expressed in online reviews, particularly on e-commerce platforms like Amazon. This study presents an automated sentiment classification model employing Bidirectional Encoder Representations from Transformers (BERT), Long Short-Term Memory (LSTM) and Bidirectional LSTM models. The model accurately extracts sentiment polarity from a dataset of 50,000 Amazon product reviews, labeled with positive, negative, or neutral sentiment. Evaluation metrics including accuracy, precision, recall, and F1-score assess the model’s performance. The Bert model achieves 91% accuracy, LSTM model achieves 88% accuracy, while the bidirectional LSTM model outperforms with 90.7% accuracy. This research provides valuable insights into consumer sentiments, enhancing decision-making processes and the online shopping experience.