AbstractWith the utility infrastructure domain becoming more technologically advanced with the modeling of all types of information across varying sectors, it is imperative to develop a domain ontology to enable the interoperability across the heterogeneous landscape of information modeling. This paper develops an ontology for the utility infrastructure domain by coupling the semantics of City Geography Markup Language (CityGML) Utility Network application domain extension (ADE)—a candidate open standard for modeling utilities—and domain glossaries, lists of utility terms with their textual definitions. First, a base ontology is formalized by abstracting the modeling information in the ADE through a series of semantic mappings. Second, a novel integrated natural language processing (NLP) approach is devised to automatically learn the semantics from the glossaries. The learning process includes the extraction of utility product terms using conditional random field (CRF) and the classification of semantic relationships between the terms using long short-term memory (LSTM) networks. Finally, the semantics learned from the glossaries are incorporated into the base ontology to result in a domain ontology for utility infrastructure. The NLP approach was evaluated using human-annotated test sets, and results show an average accuracy of 96% in term extraction and 86% in semantic relationship classification. For the case demonstration, a glossary of water product terms was learned to enrich the base ontology and the resulting ontology was evaluated to be an accurate, sufficient, and shared conceptualization of the domain. The newly developed ontology is expected to function effectively as an interoperability facilitator for the utility infrastructure domain, attributed to the semantic compatibility with existing utility modeling initiatives and the enriched/expandable semantic vocabulary.