G-Pro bot: a GraphRAG-powered GIS (Geographical information system) chatbot

Generative artificial intelligence (GAI) is gaining increasing popularity, but its applications in the Geographical Information System (GIS) domain remain limited. Consequently, answers to domain-specific questions often lack depth and specialization. GraphRAG presents a promising solution by building a knowledge graph, integrating contextual information from a knowledge base, and generating higher-quality answers. In this study, we use GraphRAG to enhance a pretrained large language model (LLM) through a retrieval-augmented generation framework, presenting a domain-specific chatbot named G-Pro Bot. Our evaluation, which includes both automated and manual evaluation, confirms the strong performance of G-Pro Bot in comprehending GIS knowledge and generating accurate, context-aware responses compared with baseline models. This makes the G-Pro Bot well suited for use in GIS education and decision-support scenarios. To the best of our knowledge, this research represents one of the earliest applications of GraphRAG in the development of a GIS chatbot, and it validates a lightweight, resource-accessible strategy for building high-performing, domain-specific artificial intelligence (AI) systems. The findings provide a valuable blueprint for creating reliable and easily deployable question-answering systems in other professional fields.