Chatbot for E-Commerce Service Based on Rasa Framework and BERT Model
Thanh-Truc Nguyen, Vu Tran, Ninh Nguyen, Trinh Huynh, Dung Dinh · 2024
The rapid growth in demand for instant and personalized responses within the e-commerce sector, particularly in online clothing retail, presents a significant challenge: delivering timely customer support to enhance satisfaction. Traditional methods, such as call centers and email support, often prove slow and costly for e-commerce businesses. Chatbots offer a practical solution by providing instant and efficient support. Among available chatbot frameworks, the Rasa Framework stands out for its simplicity and effectiveness in real-world applications. This research focuses on developing and implementing a chatbot for intelligent and convenient ecommerce services using the Rasa Framework. In our evaluation, we conducted a comparative analysis of various machine learning models. The results highligh the superior performance of Google's BERT language model, particularly when integrated with the DIET and Fallback classifiers, achieving an accuracy of 0.64, compared to 0.56 without BERT. The system was tested with 400 questions across 50 topics, categorized into three primary question groups. Approximately 64 % of the responses were correct, with users reporting relatively positive satisfaction ratings. These findings suggest that BERT-enhanced chatbots have significant potential to improve customer service efficiency in e-commerce.