Seyeon Jeong, Beakcheol Jang, Journal of Internet Computing and Services, Vol. 26, No. 2, pp. 73-84, Apr. 2025
10.7472/jksii.2025.26.2.73, Full Text:  HTML
Keywords: Agent, Large Language Model, infectious disease, Retrieval Augmented Generation

Abstract

The COVID-19 pandemic has deepened global awareness of the importance of infectious diseases and public health and the need for rapid response. To address these challenges, artificial intelligence technologies are being actively adopted, but existing models rely on trained data and are unable to effectively respond to emerging infectious diseases and untrained data. To this end, this paper aims to design and build an infectious disease AI agent that can provide richer infectious disease information than a self-contained LLM and respond to the latest infectious disease information without fine-tuning by utilizing a Large Language Model. The infectious disease AI agent is built on a Retrieval Augmentation Generation system and web search API based on a large language model. In this study, we collected and utilized domestically produced infectious disease data. Experiments on QA data showed a performance improvement of 4.44% on precision in BERTScore and 8.59% on sentence similarity evaluation. Additionally, the GPT-4o mini-based hallucination score evaluation showed a 36.8% improvement over the previous system. This research suggests that infectious disease AI agents have the potential to become an important tool in public health crisis management, and suggests a wide range of possible applications.

Statistics
Cite this article