Vector data is a core structure in geographic information science (GIS), with Shapefile being the industry standard format. However, processing such data requires domain expertise, obstructing interdisciplinary research in spatial data analysis. While large language models (LLMs) exhibit breakthroughs in natural language processing and task automation, they still face challenges in handling the spatial-topological complexities inherent in GIS vector data. To address these challenges, we propose ShapefileGPT, an innovative LLM-powered framework for automating Shapefile processing in spatial analysis. ShapefileGPT utilizes a multi-agent architecture comprising the planner and worker. The planner orchestrates task decomposition and supervision, while the worker implements spatial operation execution. We organized a custom spatial analysis library with API documentation, enabling the worker agent to efficiently process Shapefiles through function calls and validate the framework’s effectiveness. For evaluation, we constructed a preliminary evaluation dataset based on authoritative textbooks, encompassing tasks in categories such as geometric operations and spatial queries. ShapefileGPT achieved a 95.24% task success rate, outperforming GPT models. Compared to general-purpose LLMs, it effectively handles complex vector data analysis, demonstrating superior spatial data understanding and analytical capabilities. This innovation expands the possibilities for GIS automation, with promising implications for interdisciplinary spatial analysis and applications.
Lin et al. (Fri,) studied this question.