A Novel Chatbot Driven by Sentiment Analysis Using a Capsule Network and BiLSTM for Online User Feedback on Consumer Electronics

The rapid advancements in Internet use have necessitated a growing demand for data analytics to analyze customer and seller activity on consumer electronics e-commerce platforms. Online purchasing has transformed lifestyles, shopping habits, and consumer gadgets. Neglecting aspects such as customer acquisition, retargeting, answering inquiries, and inventory management can negatively impact e-commerce businesses. Online merchants like Amazon and Flipkart utilize artificial intelligence (AI) to estimate customer purchases. AIdriven patterns introduce chatbot-based and personalized suggestions to better understand customer behavior. To forecast and identify the right customers, these leading industries analyze millions of buyer interactions. This research develops a retrievalbased chatbot using sentiment analysis (SA) for consumer electronics which collects user input, that helps improve service and refine customer feedback. The effectiveness of text processing is evaluated using Term Frequency-Inverse Document Frequency (TF-IDF) and word2vec-based vectorization techniques to extract and select features. Developed features, along with LIWC-22-based characteristics, are used to categorize feedback as positive or negative. The classification is performed using the Capsule Network and Bidirectional Long Short-Term Memory (CapsNet-BiLSTM) model, as well as the XGBoost model. This study has achieved the highest scores of 98.39% for accuracy, 98.81% for precision, 98.01% for recall, and 98.41% for the F1-score, the results demonstrate that the suggested method is highly accurate and delivers superior performance. Simulations confirm that the proposed consumer behavior forecasting approach outperforms existing methods.