Open cmungall opened 6 years ago
Just a random thought, but I wonder if the inverse issue has to do with a "manually curated vs. fully-inferenced" version issue. In other words, is Ontobee sucking in the version that contains all the inference-generated info? Because if you assert the inverse only in one direction, the reasoner should conclude the inverse in the other, right?
For OBO library ontologies, Ontobee downloads the ontology owl files from the websites whose URLs are recorded in the obofoundry inventory, and processes and stores the owl file RDF results into a RDF triple store, and then displays in the Ontobee website. We did not do an additional reasoning/inference in Ontobee. It was suggested by some that we better do an additional reasoning. But some others suggest that we do not need to do it. I guess that this can still be discussed.
I would not pose this as a reasoning problem. In fact pre-reasoning in OWL would not help, DisjointClasses takes a set as argument, so Disjoint(A B)
and Disjoint(B A)
are structurally equivalent as well as semantically.
You could pose this as an RDF-level reasoning problem but I would just state as a programmatic requirement: show disjointFrom in both directions.
btw we will likely incude more inferences in released RO when ROBOT supports, eg domain/range propagated down. But as I say this doesn't help with DisjointClasses, EquivalentProperties and other constructs that takes sets as arguments
Good point. Let's see how Ontobee can be updated to address this. Edison will be working on it.
Could be done via a simple SPARQL UPDATE rather. This is something that might be generally useful for multiple triplestores, or for ROBOt querying etc, cc @balhoff @jamesaoverton
Looks like @e4ong1031 has addressed the inverseOf issue in the source code: #137
@cmungall Yes, the inverseOf was addressed in the pull request #137 . Also, the definition source URL was also addressed in #136 .
Releases are not kept up to date
Last release was may https://github.com/oborel/obo-relations/releases
ontobee version here is March http://www.ontobee.org/ontology/RO
Definition source URLs are turned into numbers
Compare http://purl.obolibrary.org/obo/RO_0002507
Protege:
In ontobee this just shows up as a number
inverses only show up in one direction
These are mutual inverses:
http://purl.obolibrary.org/obo/RO_0002508 http://purl.obolibrary.org/obo/RO_0002507
but the inverse only shows for one direction
This is particularly important for RO because we tend not to duplicate redundant info on the inverse. We choose a 'main' direction and then include a pointer