A dump of ML and ML-related RDF/OWL ontologies because some of them are impossible to find and the semantic web community can't get its links together.
Applicability score / 10: how useful it is for practical work
Coverage score / 10: how complete it seems
Both of these scores are ad-hoc and focused on answering questions about ML work; if the ontology seems like it could not answer such questions, its applicability will suffer
| Name | Applicability / 10 | Coverage / 10 | Comments |
|---|---|---|---|
| AI4EU | 2 | 2 | way too small, uber-simplistic |
| ANNETT-O | 7 | 7 | probably the most focused on ANNs, has a lot of good info there, but of course, no re-use of anything else, all stovepiped |
| ANNO | 7 | 5 | like ANNETT-O but uglier, not as complete, imports a few old ontologies but none related to AI |
| ed-ai | 3 | 3 | very simplistic, seems business-focused, more mid-level, perhaps even high-level, imports nothing, all stovepiped |
| mex (all three) | 7 | 8 | decent ontology suite, all stovepiped except small usage of DOAP |
| ML-Schema | 7 | 10 | we all know ml-schema, very general and technically useful as a high-level binding, but has zero re-use and isn't connected to anything higher |
| nno | 5 | 5 | a simpler mex / anno, pretty clean, no re-use of anything |
| Onto-DM | 9 | 9 | probably the best bet if you have the will to understand it, re-uses BFO so it is interoperable but also imports all of OBI which means half of the taxonomy is littered with irrelevant biomedical terms, definitely packed with good knowledge |
Onto-DM is the best bet, but unfortunately it is old and dead. It is nearly impossible to located on the internet and documentation is atrocious. It is of the most interest because of its adherence to BFO.