Enabling Natural Language Access to BIM Models with AI and Knowledge Graphs

Ibba A.;Alonso R.;reforgiato Recupero D.

2025-01-01

---

Abstract

Building Information Modeling (BIM) centralizes project data within a unified digital framework, enhancing collaboration across the Architecture, Engineering, Construction, and Operation (AECO) sector stakeholders. However, querying BIM data remains challenging due to the complexity of formats such as Industry Foundation Classes (IFC), which require specialized expertise. Existing tools provide limited functionality when attempting to extract information through natural language interactions, while Large Language Models (LLMs) struggle with IFC data due to its scale and complex relationships. The proposed approach addresses these limitations by integrating LLMs and knowledge graphs (KGs) to facilitate natural language queries. By structuring BIM data as a KG prior to LLM processing, we are able to enhance the extraction of knowledge while preserving semantic integrity. Evaluated on a multi-storey building, our approach demonstrates the potential of graph-based AI for BIM analysis.