Matthew Asato
Research Mentor(s): Yongqun He
Mentor Department:
Authors: Matthew Asato, Feng-Yu (Leo) Yeh, Yongqun Oliver He, Jie Zheng
Session: Session 7 (4:00pm – 4: 50pm)
Presentation Type: Poster 4
Abstract
In this UROP project, we developed VaxChat, a program designed to increase the accessibility of accurate vaccine information by leveraging LLMs to convert natural language to domain-specific querying language. It provides users with accurate answers utilizing Retrieval Augmented Generation (RAG) to reduce hallucination and ground answers in data. It connects to VaxKG, a knowledge graph integrating two He Lab databases—Vaccine Investigation and Online Information Network (VIOLIN) and Vaccine Ontology (VO)—to enhance data connectivity and relationship representation. Our evaluation has shown that VaxChat worked effectively and have achieved our designed features.
