Towards a Barrier-free GeoQA Portal: Natural Language Interaction with Geospatial Data Using Multi-Agent LLMs and Semantic Search

TL;DR

The paper tackles the barrier for non-experts to access geospatial data by introducing a barrier-free GeoQA portal that leverages a multi-agent LLM framework to decompose natural language questions into explicit geospatial subtasks. It combines semantic region selection, hybrid entity retrieval over a vector-augmented knowledge graph, and in-memory spatial analytics with transparent task planning and interactive visualizations. Key contributions include the Analyzer/Explainer/Visualizer architecture, a Region Selector using Nominatim, a hybrid Entity Retriever with vector and graph-based retrieval, and an Explainer that generates charts and Cypher queries to promote transparency, validated through case studies and user testing showing improved accuracy and usability over baselines. The approach enables flexible data input, cross-database integration, and barrier-free interaction, with demonstrated potential to improve public access to geospatial information and inform future geoportal designs.

Abstract

A Barrier-Free GeoQA Portal: Enhancing Geospatial Data Accessibility with a Multi-Agent LLM Framework Geoportals are vital for accessing and analyzing geospatial data, promoting open spatial data sharing and online geo-information management. Designed with GIS-like interaction and layered visualization, they often challenge non-expert users with complex functionalities and overlapping layers that obscure spatial relationships. We propose a GeoQA Portal using a multi-agent Large Language Model framework for seamless natural language interaction with geospatial data. Complex queries are broken into subtasks handled by specialized agents, retrieving relevant geographic data efficiently. Task plans are shown to users, boosting transparency. The portal supports default and custom data inputs for flexibility. Semantic search via word vector similarity aids data retrieval despite imperfect terms. Case studies, evaluations, and user tests confirm its effectiveness for non-experts, bridging GIS complexity and public access, and offering an intuitive solution for future geoportals.

Figures (16)

• Figure 1: Overview of the proposed framework. It contains three main components: (1) Analyzer, (2) Explainer, and (3) Visualizer. Each component would include LLM agents or pre-implemented functions. The framework provides three possible outputs for users: explanations, reasoning procedures, and visualizations. Agents handle dynamic reasoning, while pre-implemented functions perform specific computational tasks. Different colors in the figure distinguish key processing units.
• Figure 2: Workflow of the module Region Selector.
• Figure 3: Relationships between databases in a graph structure. The green nodes indicate database types, while the red nodes correspond to table or graph names. Brown nodes represent column or type names categorized as name, and blue nodes represent column or type names categorized as category. Through these connections, the matched text can be mapped to the corresponding entity information within the database.
• Figure 4: Workflow of the module Entity Retriever.
• Figure 5: Workflow of the module Data Analyzer.