OGD4All: A Framework for Accessible Interaction with Geospatial Open Government Data Based on Large Language Models

2025-11-30

Conference on Algebraic Informatics

AI Summary

This work proposes a transparent, auditable, and reproducible framework to address the high barrier to querying geospatial open government data and the susceptibility of large language models (LLMs) to hallucination. The framework uniquely integrates agent-based reasoning with secure, sandboxed code execution: it employs semantic retrieval to identify relevant datasets, enables an intelligent agent to generate verifiable code that is executed within a sandboxed environment, and produces multimodal outputs. Crucially, the system reliably abstains from answering queries unsupported by available data, substantially mitigating hallucination risks. Evaluated on a benchmark comprising 199 questions, 430 datasets from the City of Zurich, and 11 LLMs, the system achieves 98% analytical accuracy and 94% recall while ensuring interpretability and societally relevant outputs.

Abstract

We present OGD4All, a transparent, auditable, and reproducible framework based on Large Language Models (LLMs) to enhance citizens’ interaction with geospatial Open Government Data (OGD). The system combines semantic data retrieval, agentic reasoning for iterative code generation, and secure sandboxed execution that produces verifiable multimodal outputs. Evaluated on a 199-question benchmark covering both factual and unanswerable questions, across 430 City-of-Zurich datasets and 11 LLMs, OGD4All reaches 98% analytical correctness and 94% recall while reliably rejecting questions unsupported by available data, which minimizes hallucination risks. Statistical robustness tests, as well as expert feedback, show reliability and social relevance. The proposed approach shows how LLMs can provide explainable, multimodal access to public data, advancing trustworthy AI for open governance.

Problem

• Open Government Data
• Geospatial Data
• Large Language Models
• Accessible Interaction
• Trustworthy AI

Innovation

• Large Language Models
• Open Government Data
• Geospatial Data
• Agentic Reasoning
• Multimodal Output