Meeting minutes
RDF lists
jim: Working on the change to OWL API to parse ontologies that use RDF lists, to enable it to work in Protege.
… Got it working. Not too many changes. It parses into property assertion axioms correctly, but ran into something more tricky.
… If you annotate one of these axioms (like reification) there's some weird things that happen if you have an RDF list with an annnotation on it.
… It gets transferred to other axioms. The parsing gets confused. Still need to figure that out.
Concept IRIs
david: A couple of people (Jim and I) pushed back on restricting the last char of the stemIRI.
eric: I agree.
jim: They retroactively changed JSON-LD so that when you have a property in the @context, it can only act as a prefix if it ends in a hash or slash. In JSON-LD 1.1 there's a new keyword to allow other expansion.
david: if we're using JSON-LD 1.1 to do this expansion, would it be a compatibilty problem if we don't restrict the last char of the stemIRI?
jim: We might need to include that special declaration.
eric: If there's a stemIRI that someone wants to use, that did not end in slash/hash/etc, then if we allow that, and somebody wanted to use that same stemIRI and use it as a prefix in JSON-LD at the root.
https://
"APO": {
"@id": "http://purl.obolibrary.org/obo/APO_",
"@prefix": true
},
jim: Here's the issue where Harold ran into this problem: https://
david: So the @context generator needs to have a test, and emit an '@prefix: true' if the stemIRI does not end with slash or hash.
eric: The preprocessor adds type arcs for Codings. Looks at the system, appends the code, and adds a new property to the JSON (for the type arc). It can emit that type arc either using a prefix or using an absolute IRI.
david: But where would it get the prefix, to emit it?
eric: Good point. Could just emit ns1:, ns2:, etc.
eric: I prefer to be able to read the JSON-LD @context if it distinguishes between prefixes and properties.
david: So that means emitting '@prefix: true' for the stem IRIs? Eric: Yes.
eric: And warn in a Security section about choosing stemIRIs that could form unexpected domain names when the percent-encoded code is appended.
jim: Actually the 1.1 limits the last char to more than hash and slash: they are all gen-delims, as defined in RFC3986 or 3987.
https://
guoqian: In OWL API and protege, do you have use cases to demo inference using RDF data?
jim: I looked at Harold's tutorial. I use OWL and protege quite a bit, such as with Gene Ont.
https://
jim: We do inference on Signal transduction pathways, regulating processes in cells, but not using FHIR data.
ACTION: Jim to provide documents to Guoqian on use of inference
eric: We also have old examples using OBO stuff, but not using FHIR.
https://
https://
guoqian: FHIR RDF uses a lot of blank nodes. Can the rendering be improved in protege?
jim: RDF lists hide the list content, when looking in protege.
eric: if I pass it a graph, and ask for things that are in a subclass, it will get them even if they're in an RDF list, right?
jim: yes.
david: I think this points out an unfortunate bifurcation that has happened in the Sem Web community: some in the OWL community seem to view RDF as separate from OWL, rather than being the basis for OWL, and this has caused some lack of RDF support in OWL.
guoqian: Want to consider using FHIR RDF data w SNOMED ont. Harold showed a simple example. Want to scale that more.
david: Clinical Decision Support might be a rich source of more use cases. Ask Claude?
eric: I ran into tooling problems when I wanted to mix SPARQL and OWL, but the only SPARQL interfaces would only query either Abox or Tbox, but not both.
… Might need to pay for stardog to do it.
david: Did the reasoning that you wanted require backward chaining or only forward?
eric: Non-viral respiratory infections.
… i guess you could do it in RL.
david: Sounds like you could instantiate all of the inferences and then use regular SPARQL.
jim: That's what I normally do.
ADJOURNED