Addresses represent essential information used across society, but accessing them accurately and comprehensively remains challenging due to complex legal structures, distributed information sources, and emerging formats such as object and three-dimens...
Addresses represent essential information used across society, but accessing them accurately and comprehensively remains challenging due to complex legal structures, distributed information sources, and emerging formats such as object and three-dimensional addresses. To address these issues, this study proposes a domain-specific Agentic RAG-based chatbot designed to enhance accessibility to address information. The system leverages authoritative resources, including the Road Name Address Act and official administrative manuals, and performs step-wise reasoning and evidence verification according to query complexity, ensuring accuracy and legal consistency. Quantitative and qualitative evaluation results indicate that the proposed system achieves approximately 8.72% to 17.36% improvement compared to existing approaches. This research has the potential to reduce information-access gaps and improve work efficiency in the address domain.