Requirements Engineering for LLM: The Case of Digital Inquiries Application

Communication between family physicians (FPs) and patients via digital inquiry systems has become a widespread practice, often replacing in-person meetings and phone calls. Studies around the world have shown that while digital communication is deemed very important, it often results in FPs frustration and burnout as well as patient dissatisfaction. This study followed our previous study that examined and mapped existing types of inquiries and typical replies to such. This paper reports on the tasks associated with requirements engineering before the building of the Large Language Model (LLM) designed to automatically identify urgent requests for physicians and to route the request to relevant administrators or nurses if deemed necessary. In cases where the request is inappropriate for digital communication, the patient will be guided to schedule a phone call or in-person visit to the physician. The paper presents the LLM requirements engineering process as conducted in the current study and suggests a methodology for building LLM in general.