I just looked at the detail of the DATS example and it has loads of subject-specific dataset properties that can't be classified as generic dataset-version properties; basically everything that is currently commented out in the committed example. Can we use a relation to a generic entity for such cases? Would #102 be the approach to follow?

E.g.:

"isAbout": [
        {
              "identifier": {
                        "identifier": "https://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?id=9606",
                                        "identifierSource": "NCBI Taxonomy Database"
              },
              "name":"Homo sapiens"
        },
        {
            "name": "adult"
        }
    ],

Conceptually: how would we include the triple: <dataset-version> <isAbout> <homo sapiens> using the appropriate (c)uri and appropriate (existing?) classes in our schema?

As long as the object in the triple is an Entity the qualified relation pattern should be applicable.

Your example above appears to be a topic relation https://patterns.dataincubator.org/book/topic-relation.html

DCAT also has this built-in.

So what we can do is to use the existing pattern: a place to define the relation entity (relation), and then give it a role via qualified_relation.

Alternatively, we can support primary_topic. Not sure, yet.

The more complex case is when we have no property defined, and the annotation is just a data property (like an integer). 8b39663 proposes an approach for that.

I will merge this now. Before completing the example, I want to write up a document that explains the general ideas (to me).

I feel like there is a danger to become needlessly complex, and we should sharpen the focus a bit to better be able to tell what we need and what not.