Semantic enrichment
Automatically adding implicit meaning — classification, function, topology — to a model with rules or machine learning.
Semantic enrichment is the automatic or semi-automatic addition of meaningful, domain-specific information to a digital model — inferring new data from what is already explicit and implicit in the model, using logical rules or machine learning. It is the data-driven counterpart to modelling: making a model mean more, not just show more.
What it adds is the information latent in the geometry. Object classification (recognising a load-bearing wall behind a "generic mass"), the function of spaces (telling a lecture hall from a corridor by geometry and topology), and the topological relationships between elements. Methods range from rule-based "if-then" inference engines to neural networks for image recognition, to semantic-web ontologies (ifcOWL).
It matters because it closes the gap between rich geometry and poor semantics. It enables automated code-checking — regulatory model checking needs semantic data that models often lack — supports data-driven workflows (analytics, digital twins), and works around the limits of IFC exchange, which is too generic to carry full meaning on its own. It is an emerging research field — formalised by Belsky, Sacks and Brilakis (2016) and reviewed by Bloch (2022) — and is especially valuable for giving meaning to survey-derived models such as as-built and heritage models. In Italian it is arricchimento semantico.
Need this in practice?
Information Modelling →