BABot: A Framework for the LLM-Based Chatbot Supporting Business Analytics in e-Commerce

In recent times, there has been a surge in the use of chatbots built from Large Language Models (LLMs). By leveraging text generation capabilities instead of being confined to a predefined script, LLM-based chatbots have gained popularity and exerted a significant influence on the development of chatbot technology. This paper presents a framework for a chatbot that conducts conversations about business analytics in e-commerce. The framework is based on GPT-4, an LLM model capable of generating responses to assist users in deriving business insights from data and making informed decisions. The proposed framework is constructed using a process that combines several techniques for data aggregation, data visualization, and insight generation incorporating with various prompt templates. It is implemented by integrating into Trivi, a customer-intelligent platform for SMEs. An experiment was carried out to evaluate its performance, using different datasets, revealing that this approach is slightly better than previous works.