Dynamic Web Scraping and Hybrid Lexicon-Based SentimentAnalytics for Apparel Products

Abstract

Modern online fashion shopping systems generate large amounts of unstructured customer feedback about clothing quality, fabric feel, fit, and comfort. This feedback grows faster than it can be analyzed manually. Most traditional tools rely on simple star ratings or numerical scores, which fail to capture detailed customer opinions. For example, they cannot distinguish between dissatisfaction with a product and complaints about shipping delays. This paper introduces the Product Sentiment Analyzer and Review Dashboard, a system that automatically collects and analyzes customer reviews. The system uses a dynamic browser automation layer powered by Selenium to retrieve real-time feedback from product pages on major e-commerce platforms such as Amazon and Flipkart. Since these pages load content dynamically using JavaScript, the tool captures complete user reviews. After collection, the data is processed using a hybrid natural language processing approach that combines TextBlob and VADER models to classify reviews as positive, negative, or neutral. The processed data is stored in either SQLite or MongoDB Atlas databases and served through the Flask web framework. Results are presented using an interactive interface built with React.js and Chart.js, enabling clear visualization of customer sentiment. Experimental evaluation demonstrates that the system efficiently retrieves data without being blocked, provides fast database responses, and accurately classifies customer opinions. These capabilities make it an effective solution for supporting informed purchasing decisions and enabling real-time brand reputation management in the fashion industry.