Product capability · Semantic Intelligence
Build one governed ontology, not another inconsistent glossary.
Construct one evidence-qualified ontology
Data owners need reusable definitions that can be traced back to actual sources. OntoGuard automates evidence-qualified construction, preserves different senses and keeps gaps visible.
Why it matters
Reduce manual mapping and review effort; inspect each concept’s definition, domain, category, evidence and standing rather than treating generated terms as truth.
Source to constituted concept
Deterministic, native-first construction distinguishes a raw string or field from a constituted ConceptSense. Exact source locators support definitions and domains. Optional bounded provider assistance may supply evidence but does not govern simply because a model produced an answer.
One ontology, not a preferred subset
Each constituted visible concept has identity, populated evidence-grounded Definition and Domain, Concept Category and semantic standing. Categories include business, technical, structural, scientific, legal, operational, property-like, event-like and measurement-like. Class and Role are orthogonal dimensions and may be unknown. Needs Review remains visible within the one constituted ontology. In the controlled rich run, 22 visible constituted concepts include two review-required rows; the native accepted/counting summary is 20. Review standing is not a separate primary ontology.
General Ontology organizes concepts; the phenomenological layer records evidence and observations; selective micro-ontologies add specialized meaning. These are complementary layers, not separate products.
Open-world UNKNOWN is not false. Conflicting evidence remains visible, with alternatives and review requirements rather than silent promotion.
Evaluate the runtime
Define the source or protected-action boundary, success criteria and integration responsibilities before a focused evaluation. No confidential data through the public site.
Evidence and buyer questions
Can it build an ontology from a dictionary?
Yes, where evidence supports constitution. The v1.89.332 rich controlled run from two synthetic workbooks has 22 visible constituted concepts, 20 candidate edges, 45 exact source locators and two Needs Review rows. Unsupported observations do not automatically become concepts.