A Human-like Interactive Chatbot Framework for Vietnamese Banking Domain

In recent years, the application of chatbots evolved rapidly in numerous fields and received increasing attention in the academic and industrial communities. In this paper, we present a novel chatbot framework based on machine learning and deep learning approaches. Our framework not only answers the domain questions but also consists of three primary features of a human-like interactive chatbot, including (1) Conversation tracking, (2) Recommendation, and (3) Asking again. Further-more, we integrate one feature for adding accents to the non-accent sentence using Transformer-based architecture. Based on the experimental results and deployment on production for the banking domain, we demonstrated that our framework is stable and ensures specific requirements (e.g., computational resources, response time, performance, user experience). With flexibility and adaptation, our proposed framework can be developed and deployed to other domains or business contexts.