coleridge-initiative / adrf-onto Goto Github PK

• You can't just graft classes from different ontologies, instantiate them and put them into a SKOS concept scheme
• Eg when you say :Dataset a dcat:Dataset, it means that :Dataset is one particular dataset. This cannot be owl:sameAs dbr:Data_set because that's (arguably) the concept of a Dataset.
• Eg when you say :Provider a foaf:Organization, it means :Provider is some specific organization. And it's doubly impossible to say owl:sameAs dbr:Data_publishing because publishing (providing) data is an activity, not an organization.
• prefix : is missing, so it seems to me https://github.com/Coleridge-Initiative/adrf-onto/blob/master/adrf.ttl is invalid
• https://github.com/Coleridge-Initiative/adrf-onto/wiki/Vocabulary says "subclassed from" but a (rdf:type) is not subclassing, it's instantiation
• It says "Corpus Subclassed from SKOS:Collection" (even though that's missing from the actual ontology), but skos:Collection has a specific use (ad-hoc collection of Concepts), whereas Corpus is a collection of documents (texts), so you can't do that
• you have two different definitions of :Topic. One is tied to LCSH, which is wrong

It seems to me you are in dire need of some ontology consulting, by someone who knows ontologies for describing:

• machine learning and data science (there are at least 8 I know of)
• datasets (DCAT2 and ADMS are the most important)
• corpora and NLP (NIF and friends)

Continuing the examination from #3:

• again the : prefix is missing, so rcc.ttl is invalid
• what's the purpose of using xsd:anyURI literals rather than actual URIs?
• You say :Catalog :dataset :dataset481, :dataset_x001 but you have not defined such property. Thinking about dcat: doesn't magically bring its props into your namespace
• You say :Corpus :publication :publication338, :publication340 but you have not defined such property

We've done significant work in KGs of Science so the following recommendations come from that experience:

• authors and publications: use separate props to capture identifiers in important datasets like ORCID, DOI, Google Scholar, ResearchGate, Scopus. Eg Wikidata has external-ids for all of that
• publications: use some established bibliographic ontology, don't just wing it with dc/dct
• If you use DCT for a bunch of props, why suddenly switch to PAV for pav:createdOn? Use dct:created
• dct:alternative is declared subprop of dct:title, so you'll get two dct:title: "National Health and Nutrition Examination Survey" and "NHANES", is that what you want?