Ontologies form the basis of the knowledge model and therefore are critical to proper functioning of the knowledge graph. “One way of implementing an ontology is to start with a prebuilt industry ontology, which exists for many verticals, including life sciences and financial services,” said Andreas Blumauer, co-founder and senior VP of growth at Graphwise. “They can then be populated with enterprise-specific information.” The other way is to bootstrap from the beginning. “If nothing is available, then an ontology can be built from existing documents, taxonomies, and classes of activities,” he added.
Starting with a tool such as the Graphwise Taxonomy Builder, which is a feature of Graph Modeling, a rough outline of how the domain should be described can be developed. Then a large language model (LLM) can add information, followed up by enterprise information. “It is an interactive process,” noted Blumauer, “and should be done in conjunction with domain experts.”
A small but carefully selected set of nodes forms the inner core of a knowledge graph, and then the data gets linked in to form a full-fledged knowledge graph. “There is a recipe of how you should integrate your data from the silos,” Blumauer explained. “If you don’t have good domain knowledge, you end up with just what you had in the databases, rather than enriched information.”