Clustering Information Types for Semantic Enrichment of Building Information Models to Support Automated Code Compliance Checking

Missing, incomplete, implicit, and/or incorrect information are major obstacles to automated code compliance checking in the construction industry. All existing platforms for automated code checking require users to extensively preprocess their input models to supplement missing information before checking can begin. Semantic enrichment using artificial intelligence (AI) can automate much of this normalization process. Progress in the field of semantic enrichment, in turn, requires identification and specification of the information types that must be made explicit, and of the procedures appropriate for each type. After characterizing a broad set of clauses from five diverse building codes, a two-stage clustering process with the k-means algorithm was used to derive a hierarchical classification of semantic enrichment task types. The resulting classification defines 10 tasks that are typically needed for automated code compliance checking. Future research can build on the classification to formalize a knowledge base to inform selection of appropriate approaches for semantic enrichment tasks.