Have a chat with clustine, conversational engine to query large tables

Abstract

Thanks the recent advances of AI and the stellar popularity of messaging apps (e.g., WhatsApp), chatbots are no longer bound to customer support services and computer museums. Indeed, they provide a mighty, lightweight and accessible way to provide services over the Internet. In this paper, we introduce Clustine, a chatbot to help users query large tables through short messages. The main idea is to combine cluster analysis and text generation to compress query results, describe them with natural language and make recommendations. We present the architecture of our system, demonstrate it with two use cases, and present early validation experiments with 12 real datasets to show that its promises are reachable.