Intelligent knowledge graph question answering method for health and safety hazard management using large language models

Construction safety knowledge is often scattered across various unstructured or semi-structured sources, complicating its retrieval, reasoning, and application for effective safety management. This paper introduces a novel Knowledge Graph Question Answering (KGQA) method that leverages Large Language Models (LLMs) for intelligent QA over safety hazard knowledge. The method integrates an LLM-based assistant for natural language understanding (NLU), allowing users to query a domain-specific safety knowledge graph (KG) using natural language. To evaluate its effectiveness, a natural language query (NLQ) dataset is developed and used for benchmarking QA performance across prominent LLMs. Results demonstrate optimal accuracy in retrieving safety knowledge efficiently.