2025
- Natural Language Information Retrieval from BIM Models: An LLM-Based Agentic Workflow Approach. Proc. of the European Conference on Computing in Construction (EC3), 2025 mehr…
Sylvain's PhD research is dedicated to establishing scalable agentic AI architectures for a universal answer engine tailored to property data. This overarching vision seeks to transform how information is accessed and utilized across the entire building lifecycle, from initial design and construction through ongoing facility management and maintenance.
The current research phase focuses on developing an AI-powered system capable of intelligent information retrieval from Building Information Models (BIM) in IFC format. This system leverages Large Language Models augmented with specialized tools, enabling multi-step, multi-hop queries through natural language without requiring prior data pre-processing. The developed system functions both as a direct query interface for human users and as an integrated component within larger automated systems, such as those used in facility management where contextual BIM information is systematically retrieved to support operational workflows.
Looking ahead, the next research steps aim to expand this methodology to encompass diverse document types associated with construction projects, including architectural drawings, contractual agreements, and technical specifications. This comprehensive system will provide intelligent document retrieval and automated question-answering capabilities. A key principle of this future work is maintaining transparency and verifiability through clear source attribution and document traceability, allowing users to cross-reference answers with original information sources.
This research addresses fundamental challenges in knowledge management within the built environment by developing novel approaches to information accessibility and knowledge extraction from heterogeneous data sources.
BIM.advanced (SuSo 2025)