Enhancing Personalized Service Development with Virtual Agents and Upcycling Techniques

Abstract In this study, aimed at achieving Society 5.0, we explore efficient development methods for services tailored to individual user needs. The diversification of user demographics has led to swiftly changing service demands, increasing the necessity for personalized adaptation. However, developing services tailored to individual users requires specialized knowledge and incurs high costs. This research focuses on developing technologies that enable service development through a concrete understanding of user needs. Utilizing an interactive needs extraction system with a virtual agent (VA) developed in prior studies, along with the Sharing Upcycling Cases with Context and Evaluation for Efficient Software Development (SUCCEED) system, we aim to automate the extraction of user needs and provide developers with relevant insights. Prior research has found it challenging to grasp the outline of services to be developed from user statements. However, a new method using large language models (LLM) has enabled the automatic extraction of novel service needs. Experiments have shown the potential to automatically provide development insights for 67% of ideal needs.