various comments on patterns owl file

[cmungall]

It's hard to see on the README where to start to get an overview (some of this looks like text I wrote a while ago?) Would browsing patterns.owl be a good place to start? Should we have a rendering (e.g. vowl, spec2owl, owldoc) linked from the repo?

The patterns owl is great but a few oddities:

there seem to be some duplicate labels:

Though the subclass has a URI with noParts in it - what does this mean?

When looking at pattern-with-imports.owl it seems the subclass is inheres-in and the parent is inheres-in-part-of

These two should be inferred to be equivalent, since all Ps occurs in a IC

It looks like the desired classification is only achieved through incomplete axiomatization. I suspect our patterns.owl are rife with these. It's a cute trick to infer subClassOf, but there are incorrect CWA/OWA assumptions here, I just don't think it will work in general. It may be better to simply assert is-a in the patterns.

Looks like there is an attempt to

[matentzn]

hahha, never thought about that; it's true, probably the hierarchy is going to have quite a few of these cases (occurs in independent continuant will always be true). I dont mind so much to be honest - Do we really want to assert subclass of axioms for 200 and more patterns?

Typically we look at Monarch OLS and the patterns docs to look for patterns. Is there a high marginal gain to introduce other visualisations? (there is code in the repo to generate the VOWL viz but its not that helpful IMO)

The noParts issue you uncovered is an artefact. I was always trying to keep the number of IIPOs in patterns low, but for the main grouping pattern for AEs the community insisted on IIPO. I later introduced a second "parallel" pattern for inheres in in just called it xyznoParts, but thats only because I wasnt creative on that day. The reason why they are named the same is, well, the label grammar is the same (just different relation in EQ). Confusing?

[cmungall]

I think you including a link to the the docs README from the top level docs would obviate a lot of my ticket (I'm now not to sure looking at the owl in Protege or OLS buys you that much at the moment (if we do want to encourage it we should have the same information))

Do we really want to assert subclass of axioms for 200 and more patterns

I have much to say on this topic but best saved for our calls

[cmungall]

cc @balhoff