Ensuring compliance with complex industry standards and regulations during the design and implementation phases of construction projects is a significant challenge in the building information modeling (BIM) domain. Traditional manual compliance checking methods are inefficient and error-prone, failing to meet modern engineering demands. Natural language processing (NLP) and deep learning methods have improved efficiency and accuracy in rule interpretation and compliance checking. However, these methods still require extensive manual feature engineering, large, annotated datasets, and significant computational resources. Large language models (LLMs) provide robust language understanding with minimal labeled data due to their pre-training and few-shot learning capabilities. However, their application in the AEC field is still limited by the need for fine-tuning for specific tasks, handling complex texts with nested clauses and conditional statements. This study introduces an innovative automated compliance checking framework that integrates LLM, deep learning models, and ontology knowledge models. The use of LLM is motivated by its few-shot learning capability, which significantly reduces the need for large, annotated datasets required by previous methods. Deep learning is employed to preliminarily classify regulatory texts, which further enhances the accuracy of structured information extraction by the LLM compared to directly feeding raw data into the LLM. This novel combination of deep learning and LLM significantly enhances the efficiency and accuracy of compliance checks by automating the processing of regulatory texts and reducing manual intervention. This approach is crucial for architects, engineers, project managers, and regulators, providing a scalable and adaptable solution for automated compliance in the construction industry with broad application prospects.
Over 40% of the world’s population lives within 100 km of the coast, creating an urgent need for accurate spatial data to support climate adaptation and urban planning. However, existing road-network datasets such as Overture adopt a traffic-oriented representation in which multi-lane roads are stored as multiple parallel segments. This introduces a critical problem for spatial analysis: a four-lane road represented as four separate lines is counted four times in space-syntax and centrality calculations, distorting measurements of urban spatial structure. Traditional solutions rely on manual axial-map creation by trained experts, limiting analysis to individual case studies. Here we present a unified road-centreline dataset for 2073 coastal urban areas across 110 countries, derived using a Voronoi-based extraction method with adaptive spatial partitioning that makes medial-axis centreline extraction feasible at global scale. The extraction scales approximately linearly with network size (power-law exponent 0.91), so that a typical urban area is processed in about a minute. Comprehensive quality validation confirms that all processed urban areas achieve excellent geometric-quality grades, with an average length-preservation rate of 85.7% and full coordinate-reference-system consistency; the extracted centrelines also agree with an independent authoritative reference to within a median of 2 m. The dataset transforms 67.6 million raw Overture road segments (13.19 million km) into a unified centreline network of 63.5 million segments totalling 11.04 million km, removing the duplicated length of parallel and multi-lane carriageways while preserving connectivity, and is provided in GeoParquet and GeoPackage format as a topologically consistent foundation for large-scale comparative studies in urban morphology and climate-resilience planning.