Problem
Research questions and friction points this paper is trying to address.
transportation safety
data access
spatial queries
natural language interface
public-sector AI
Broadening Access to Transportation Safety Data with Generative AI: A Schema-Grounded Framework for Spatial Natural Language Queries
🤖 AI Summary
Local agencies often struggle to effectively leverage traffic safety data due to high technical barriers. To address this challenge, this work proposes a novel natural language interface that translates user queries into reproducible spatial operations strictly conforming to authoritative database schemas. The approach innovatively decouples natural language understanding from deterministic execution by integrating large language models, semantic frame parsing, a rule-based validation layer, and typed directed acyclic graphs to enable precise querying over PostGIS. Evaluation on Massachusetts transportation datasets demonstrates that the system successfully executes all queries, with the validation layer correcting 29% of user errors. This effectively bridges the gap between the flexibility of natural language and the rigor of structured data schemas, thereby enhancing the trustworthiness of AI applications in the public sector.
📝 Abstract
Transportation safety analysis requires integrating crash records, roadway attributes, and geospatial data through GIS-based workflows, but access remains uneven across agencies and community stakeholders. Technical prerequisites create a gap between analytical tools central to safety planning and the practitioners able to use them. Local agencies, school committees, and residents may have safety concerns but limited capacity to retrieve, filter, map, and analyze relevant data. Generative AI offers a way to narrow this divide, but its public-sector use raises questions about reliability, reproducibility, and governance. This paper presents a schema-grounded natural language interface for transportation safety analysis, using a large language model (LLM) to interpret user intent while preserving deterministic, reviewable execution against an authoritative database. User queries are translated into structured semantic frames, validated by a rule-based layer, compiled into a typed directed acyclic graph of spatial operations, and executed against a PostGIS database. This bounded design separates language interpretation from deterministic execution, keeping results reproducible and schema-grounded while removing access barriers. The framework is evaluated using a statewide Massachusetts transportation safety database integrating crash records, roadway attributes, and geospatial layers including schools, bus stops, crosswalks, and municipal boundaries. All queries executed successfully; the validation layer corrects errors in 29% of evaluation queries, reflecting the gap between flexible natural language and strict schema-grounded requirements. The results suggest that combining natural language accessibility with deterministic execution is a practical direction for broadening access to transportation safety data, with implications for trustworthy AI in public-sector planning.
Innovation
Methods, ideas, or system contributions that make the work stand out.
schema-grounded AI
natural language interface
spatial query processing
transportation safety analysis
deterministic execution
🔎 Similar Papers
No similar papers found.
