Transformer-based framework for mapping client requirements to BIM

This paper presents a comprehensive automation framework that addresses the challenge of converting diverse client-authored textual requirements into accurate, information-rich Building Information Models. Traditional workflows rely on frequent iterations between clients and architects, manual interpretation of narrative requirements, and custom parametric modeling, which leads to delays and inconsistencies.

The proposed pipeline integrates advanced Natural Language Processing with BIM to automatically interpret design intent and generate corresponding BIM assemblies. A semantic translation layer connects parsed textual entities to a curated BIM model repository and transfers the derived constraints directly into the BIM authoring environment.

Tested across multiple projects, the system achieved a 92% accuracy rate in mapping client inputs to the appropriate BIM elements. By embedding this automated translation process, the approach improves requirement traceability, clarifies stakeholder intent, and supports scalable data-driven design analysis. Overall, the contribution enables AI-assisted construction automation through the seamless integration of NLP and BIM in a unified workflow.