Transformer-based approach for automated context-aware IFC-regulation semantic information alignment

• Published in: Automation in construction 2023-01, Vol.145, p.104540, Article 104540
• Main Authors: Zhang, Ruichuan, El-Gohary, Nora
• Format: Article
• Language: English
• Subjects: Automated code checking; Building codes; Building information modeling; Deep learning; Industry foundation classes; Information alignment; Transformers
• ISSN: 0926-5805; 1872-7891
• DOI: 10.1016/j.autcon.2022.104540

One of the main challenges of automated compliance checking systems is aligning the semantics of the building information models (BIMs), in Industry Foundation Classes (IFC) format, and the semantics of the regulations, in natural language, to allow for checking the compliance of the BIM with the regulations. Existing information alignment methods typically require intensive manual effort and their ability to deal with the complex regulatory concepts in the regulations is limited. To address this gap, this paper proposes a deep learning method for IFC-regulation semantic information alignment. The proposed method uses a relation classification model to relate and align the IFC and regulatory concepts. The method uses a transformer-based model and leverages the definitions of the concepts and an IFC knowledge graph to provide additional contextual information and knowledge for improved classification and alignment. The proposed method was evaluated on IFC concepts from IFC 4 and regulatory concepts from different building codes and standards. The experimental results showed good information alignment performance. •Method for semantic information alignment to support automated code checking.•Proposed method uses transformer-based relation classification model.•Classification model relates and aligns IFC and regulatory concepts.•Concept definitions and an IFC knowledge graph provide contextual information.•Achieved good alignment performance across different types of building codes.