15 lines
1.3 KiB
Markdown
15 lines
1.3 KiB
Markdown
# Rule: Ontology Detection vs Heuristics
|
|
|
|
## Summary
|
|
When detecting classes and predicates in `data/ontology/` or external ontology files, you must **read the actual ontology definitions** (e.g., RDF, OWL, TTL files) to determine if a term is a Class or a Property. Do not rely on naming heuristics (like "Capitalized means Class").
|
|
|
|
## Detail
|
|
* **Verification**: Always read the source ontology file or use a semantic lookup tool to verify the `rdf:type` of an entity.
|
|
* If `rdf:type` is `owl:Class` or `rdfs:Class`, it is a **Class**.
|
|
* If `rdf:type` is `rdf:Property`, `owl:ObjectProperty`, or `owl:DatatypeProperty`, it is a **Property**.
|
|
* **Avoid Heuristics**: Do not assume that `skos:Concept` is a class just because it looks like one (it is), or that `schema:name` is a property just because it's lowercase. Many ontologies have inconsistent naming conventions (e.g., `schema:Person` vs `foaf:Person`).
|
|
* **Strictness**: If the ontology file is not available locally, attempt to fetch it or consult authoritative documentation before guessing.
|
|
|
|
## Violation Examples
|
|
* Assuming `ex:MyTerm` is a class because it starts with an uppercase letter without checking the `.ttl` file.
|
|
* Mapping a LinkML slot to `schema:Thing` (a Class) instead of a Property because you guessed based on the name.
|