Towards a barrier-free GeoQA portal: Natural language interaction with geospatial data using multi-agent LLMs and semantic search

Geoportals have become essential tools for accessing and analyzing geospatial data, facilitating open spatial data sharing and online geo-information management. They typically follow interaction logic similar to common GIS systems and use a layered visualization structure. However, these designs can make geoportals less accessible to non-expert users, who may find complex functionalities overwhelming and struggle with overlapping data layers that obscure spatial relationships.To address these challenges, we propose a step towards a barrier-free GeoQA Portal enabling natural language interaction with geospatial data using a multi-agent Large Language Model (LLM) framework. User queries are decomposed into individual subtasks and handled by distinct LLM agents. Our portal can process complex spatial queries and extract geographic data relevant to user interests. Additionally, task plans are presented to users, improving transparency and engagement. The portal supports both default and user-provided data, flexibly addressing diverse user needs. We enhance accessibility via semantic search using word vector similarity, allowing users to retrieve relevant data even if terms or spellings are imperfect. Through case studies, system evaluation, and user testing, we demonstrate the system’s effectiveness in meeting the geospatial data access and analysis needs of non-expert users. The results highlight how our portal bridges the gap between complex GIS workflows and public data accessibility, offering a more intuitive and user-friendly solution for future geoportal development.