Smart Chatbot System for Banking using Natural Language Processing Tools

Abstract

Abstract This research work focuses on creating a Smart Chatbot System for banking that leverages Natural Language Processing Toolkit and Machine Learning. The study involves gathering and preprocessing a dataset of user queries and responses. Key words are defined for pattern matching, and a decision tree algorithm is used to train the model. The Chatbot's performance is evaluated, and the results demonstrate the successful development of a functional Chatbot with a high F1 score of 0.97, indicating 97% accuracy in understanding user queries and providing relevant responses. User testing and a comparison with an existing system reveal significant improvements achieved through the incorporation of NLP and machine learning algorithms. The Chatbot is designed with an intuitive and user-friendly interface accessible through a web application, and administrators have a separate back-end access to effectively manage the knowledge base. The research also identifies areas for future improvement, such as enhancing accuracy, incorporating system memory and anaphora resolution, and integrating a logging system. Addressing these limitations will contribute to the advancement and effectiveness of conversation agents in the banking sector.