Leveraging Graph Based Semantic Enrichment for Enhanced Automated Code-Checking

Abstract: Despite advances in BIM technology, achieving a comprehensive machine-readable design representation for Automated Code Checking (ACC) remains a labour-intensive task. Recent research suggests applying Machine Learning (ML) for ACC, but its efficacy relies on rich semantic data, often lacking in BIM models. We propose decomposing the regulatory checks into sets of semantic enrichment tasks, each tailored with the most suitable solution. We demonstrate this approach with a case study from the Israeli regulatory requirements for security rooms. We focus on two distinct semantic enrichment tasks necessary for the check by leveraging community detection algorithms and Graph Neural Networks (GNNs).