Real-Time Knowledge Retrieval for Banking Chatbots A RAG-Based Approach to Employee Assistance

Abstract

In the corporate sector, the experience of an employee is significant to the service they provide. Especially in sectors such as banking, where domain knowledge is crucial in providing efficient service in internal and external operations. However, in recent years within Sri Lanka, experienced professionals have started to migrate and change jobs due to the financial crisis. This affects organizations because they have to constantly work with new recruits. This research aims to provide a solution to this problem by finding a method to develop a chatbot using the RAG method that can provide assistance to new recruits with domain knowledge. The RAG is supposed to store knowledge, such as information about products, policies, and relevant domain knowledge as its context. Provides answers from the domain-specific knowledge rather than the general knowledge of an LLM. Furthermore, this research focuses on optimizing the RAG model in main aspects of the RAG such as query translation, retrieval, in order to provide more accurate and reliable outputs to the user. The results of the research are to build a chatbot that helps fresh recruits find knowledge to answer customer queries during their training period. In addition, the study highlights ways to improve the response of the RAG model to be accurate and relevant for the domain of the banking sector in Sri Lanka.