16:01:08 <RRSAgent> RRSAgent has joined #hcls
16:01:08 <RRSAgent> logging to http://www.w3.org/2017/01/24-hcls-irc
16:01:10 <trackbot> RRSAgent, make logs world
16:01:10 <Zakim> Zakim has joined #hcls
16:01:12 <trackbot> Zakim, this will be HCLS
16:01:12 <Zakim> ok, trackbot
16:01:12 <dbooth> Chair: David Booth
16:01:13 <trackbot> Meeting: Semantic Web Health Care and Life Sciences Interest Group Teleconference
16:01:13 <trackbot> Date: 24 January 2017
16:09:17 <hsolbrig> hsolbrig has joined #hcls
16:09:32 <hsolbrig> So, is there anyone else on the audio channel?
16:09:49 <dbooth> nobody else on yet
16:09:59 <dbooth> i'm trying to skype call you, as a separate audio check...
16:10:01 <hsolbrig> So we don't know the source of the problem
16:10:07 <hsolbrig> Ok - jas on the skype call
16:13:11 <dbooth> harold, can you talk on the hangout call?
16:13:19 <dbooth> i don't hear anything yet
16:13:20 <hsolbrig> I am talking right now
16:13:29 <hsolbrig> So from bad to worse, I guess
16:13:40 <hsolbrig> Lemme exit Skype - sometimes it holds on to stuff
16:14:13 <hsolbrig> Exited, audio working just fine.  No idea
16:14:24 <dbooth> well, sounds like reboot time for me.  i'll be back in a few minutes
16:14:37 <hsolbrig> No one else on anyway - want to do a short Skype session wrt next steps?
16:15:03 <hsolbrig> You keep showing noise - If I could only interpret the concentric circles...
16:17:29 <dbooth> Topic: fhir:concept
16:18:31 <dbooth> harold: Grahame is on board with changing fhir:concept to rdf:type.  I ran it by Lloyd, and he seemed okay with it.
16:19:14 <dbooth> ... And it works with classifying when status=final.
16:19:57 <dbooth> ... But fhir.ttl ont shows as OWL full.  EL reasoners don't like it.
16:20:22 <dbooth> ... I want to trim down the fhir.ttl to work with EL.
16:21:01 <dbooth> ... There are two goals: reasoning and validation.
16:21:31 <dbooth> dbooth: We are using Shex for validation; OWL for reasoning.
16:22:06 <dbooth> harold: We need to generate permissible values for fhir:coding for required codes.
16:22:47 <dbooth> ... The other high priority tasks: 1. slicing 2. constraints.
16:23:04 <dbooth> dbooth: thoughts about how to do them?
16:23:13 <dbooth> harold: yes, we know how to do them.
16:24:26 <dbooth> ... We're comfortable that we can represent slicing.  We've done an analysis of constraints, and 90% are info that could be represented in the FHIR idiom but too heavy wait for that fine grained.  But they're easily representable in shex.
16:25:17 <dbooth> ... Might need a path syntax in shex, but that wouldn't be until shex 3.
16:26:45 <dbooth> ... Also need to figure out how to, if we're pulling FHIR/RDF into a reasoner, it needs to have owl:Ontology header.   I'm inclined to put a parameter on the URL to add the owl:Ontology header.
16:28:47 <dbooth> ... The name of the ontology could be the URL.
16:29:23 <dbooth> dbooth: Would there be any harm in the FHIR generator always putting it in?
16:29:49 <dbooth> harold: not sure
16:31:11 <dbooth> dbooth: I don't immediately think of any harm in doing so.
16:32:47 <dbooth> harold: I could see value in aggregating multiple patients and reasoning over that.
16:33:46 <dbooth> ... We might use FHIR query to generate the RDF data.
16:34:43 <dbooth> ... We might want owl:Ontology header on the FHIR instance data.
16:35:18 <dbooth> ... Would be cool to go into Protege and fetch a clinical record by opening a URL
16:35:49 <dbooth> ... But I cannot do that because it lacks the owl:Ontology header.
16:42:27 <dbooth> Topic: Should owl:Ontology declaration be included in instance data? -- Harold
16:43:33 <dbooth> harold: possibilities: 1. mime extension to add owl:Ontology 2. URL parameter 3. say that the RDF always includes owl:Ontology
16:44:00 <dbooth> eric: I don't like to add needless junk to a document
16:44:28 <dbooth> ... Never understood why protege needs that.
16:44:46 <dbooth> harold: OWL spec says all assertions belong to an ont.
16:44:55 <dbooth> eric: I consider it a tool issue.
16:45:40 <dbooth> ... Logical behavior if you don
16:45:59 <dbooth> 't have a URL for the ont is to make one up, and protege does that.
16:46:10 <dbooth> s/'t/... 't/
16:46:38 <dbooth> ... But when  you do URL open, it doesn't do that.  It only does that if you make a new one.
16:47:05 <dbooth> harold: that's the "fix protege" approach.  But there's an advantage in the reasoner to knowing where the assertions were made.
16:47:13 <dbooth> ... Ont header serves a good purpose.
16:47:55 <dbooth> ... Proposed that the declared URL would be the URL of the instance doc.
16:48:27 <dbooth> ... And the ont version becomes the version of the FHIR resource.
16:51:03 <ericP> https://www.w3.org/2013/12/FDA-TA/tests/RenalTransplantation/subjects.ttl -> example of nontology
16:51:17 <ericP> <http://www.w3.org/2013/12/FDA-TA/tests/RenalTransplantation/subjects> a owl:Ontology ;
16:51:52 <dbooth> eric: Here's an example where I added an owl:Ontology declaration.
16:53:10 <dbooth> harold: This won't work unless it comes out of the FHIR server.  If it isn't in there, then someone will have to write a FHIR server wrapper for it, becuase the use case is too compelling.
16:53:42 <dbooth> eric: If we just have the data we don't learn much.  We'll normally need more things.
16:54:07 <dbooth> harold: It would also make sense in the instance data to say 'imports fhir.ttl'
16:55:31 <dbooth> harold: FHIR documents can have version.  They can be right in the RDF as a declaration.
16:55:59 <dbooth> ... I do that now using an editor but not so demo friendly.
16:56:50 <dbooth> ... The question: Can we do it in the confines of FHIR?  If not, we need a wrapper service to add that declaration.
16:57:19 <dbooth> eric: One other aspect: the thing you want to import will change between use cases.  Sometimes you don't want to import anything.
16:57:42 <dbooth> ... Sometimes you want all the object properties, but also a piece of SNOMED.
16:59:06 <dbooth> dbooth: owl:Ontology or owl:import statements could always be stripped out if not wanted.
16:59:16 <dbooth> ... So we should try to support the most common uses.
17:00:40 <dbooth> eric: If I load subject-amy.ttl and it does not have an owl:Ontology declaration, does it work?
17:00:58 <dbooth> harold: no.  Protege needs to connect it to an actual filename or URL.
17:02:38 <dbooth> eric: Issues: 1. faced the same question on the fhir root node -- needed to know the URL from which it was fetched.  Could use relative URLs, but doesn't work with protege but it requires ont URL to be absolute, and also the version URL.
17:03:38 <dbooth> dbooth: Probably rooted in the fact that relative URLs don't exist at the RDF level.  They only exist in certain serializations.
17:04:57 <dbooth> dbooth: extra noise for those who don't care about OWL, but is there really any harm?
17:05:01 <dbooth> eric: I don't think so.
17:07:23 <dbooth> dbooth: the case sounds pretty compelling for including these declarations.  The only downside seems to be some extra noise for those who don't care about OWL, but no real harm. But harold has identified a concrete benefit in including them.
17:08:14 <dbooth> eric: The reason for versions is usually if the meaning changes.
17:08:39 <dbooth> harold: Version is intended to indicate immutable state.
17:10:10 <dbooth> ... Most FHIR implementations support reasoning.
17:10:39 <dbooth> eric: If we put static identifiers over all of the examples -- pick a canonical URL.
17:10:50 <dbooth> harold: Would be the URL of the resource.
17:14:26 <dbooth> eric: Technical cost is putting these identifiers in there.  Could be worthwhile, but that decision is political.
17:20:39 <dbooth> eric: Didn't we try this for subject URL already?
17:20:56 <dbooth> dbooth: That was a little different.
17:24:31 <dbooth> AGREED: We'll (provisionally) try going ahead with adding the owl:Ontology owl:imports declarations to FHIR instance data
17:24:42 <dbooth> ADJOUNED
17:24:53 <dbooth> Present: EricP, David Booth, Harold Solbrig
17:25:01 <dbooth> rrsagent, draft minutes
17:25:01 <RRSAgent> I have made the request to generate http://www.w3.org/2017/01/24-hcls-minutes.html dbooth
19:33:38 <Zakim> Zakim has left #hcls