Ontological Integrity

The preservation of inherent hierarchical relationships, entity classifications, and semantic dependencies when transforming unstructured knowledge into structured representations.

When you cluster documents or entities purely by co-occurrence or community detection, you capture what appears together but lose what belongs together categorically. A relay component and its failure mode might end up in different clusters despite having a definitional relationship.

Ontological integrity asks: does the resulting structure reflect how a domain expert would organize this knowledge, or just how the data happened to cluster?

The distinction matters because multi-hop reasoning depends on traversing relationships that are categorically meaningful, not just statistically frequent.

Related: 05-atom—multi-hop-reasoning, 06-molecule—knowledge-graph-construction, 02-molecule—taxonomy-design