Artificial Intelligence-Driven Sentiment Analysis for Product Reviews in Online Retailing Platforms

Abstract

Machine learning and large language models (LLMs) are two examples of AI methods used for sentiment analysis that can efficiently categories texts as positive, negative, or neutral. Its capacity to automatically detect emotions in texts is utilized by numerous campaigns related to marketing, social media, product reviews, and hate speech. In order to categories the tone of product reviews taken from the Amazon Customer Review dataset, this study suggests a deep learning model. A combination of CNNs and LSTMs (Long Short-Term Memory) networks makes up the model. Tokenization, normalization, stop word deletion, and sequence padding are all part of the suggested systematic approach to data preprocessing in the given method. For feature extraction, it recommends the Bag-of-Words model. It may divide the data for training and testing 80/20 with this dataset, which contains 400,000 ratings across five product categories. The LSTM layer use long-term dependency modelling to enhance classification accuracy, whereas the CNN module seeks out local textual patterns. Testing shows that the CNN-LSTM model beats more traditional models like Random Forest, Naive Bayes, and independent CNN in terms of accuracy (98%), precision (98.52%), recall (98.49), and F1-score (98.49). The findings demonstrate that deep hybrid architecture is effective for sentiment analysis of large e-commerce platforms.