Semantic Web Health Care and Life Sciences Interest Group Teleconference -- 12 Apr 2016

Planning for Montreal HL7 Meeting Tue May 10, 11:00am-12:30pm

Sajjad: I'll be coming to Montreal meeting, and will also be in the FHIR Connectathon
... Which Connectathon tracks would you recommend?

lloyd: Newcomers are welcome, including those who have only browsed the spec

sajjad: Recommendation for starter on FHIR?

lloyd: Suggest the patient track
... There are also other tutorials through the week.

dbooth: Other planning for Tuesday meeting in Montreal?

sajjad: Need to register for Tuesday?

dbooth: Yes, on HL7 site.
... Meeting room is the room for the ITS group -- they host the RDF subgroup.
... Status update, and review work products.

RDF page

eric: Updated to work with the RDF that Grahame generates

http://w3c.github.io/hcls-fhir-rdf/spec/rdf.html

Here is the previous version: http://w3c.github.io/hcls-fhir-rdf/spec/rdf-SBS.html

eric: Need a document that reflects WG decisions and has ShEx. Also need a user-facing doc.
... The RDF templates on the ShEx side should reflect the WG decisions.
... The RDF tab should capture it in a template.

dbooth: We have not previously discussed a second document, internal-facing, to catalog our decisions.
... Concerned to add only additional document that we really need.

<ericP> CodedableConcept

eric: We could put that in the issues list. Would it capture it well enough?

tony: Hard to capture the low level decisions in an issues list. Too many of them. Many of these are just the way that we have decided to write the RDF.
... Need to do a detailed review.

eric: Bunch of decisions to make, such as CodableConcept, that will be reflected in the individual resource pages.
... Those will be represented in RDF templates.
... There's the question of how to visually parse an RDF template and generate FHIR instance. That's captured in rdf.html
... Also, if our job is to write core FHIR resources, how do we write valuesets?
... If we apply only the default mapping we won't get something computable.

tony: The structure def to schema is not 1-1. We're doing the equivalent of that for OWL.

dbooth: Ready to review eric's draft?

eric: Need to work through more of the datatypes, figure out their RDF representation, to drive the generation of the RDF.
... What should we do when we have an instantiation of another datatype? E.g., an age.
... Person.age? Or just an integer?

harold: Do we make the range integer, or create a specialized type for age?

eric: Maybe look at SBS document to see what needs to be resolved, to make a TODO list: http://w3c.github.io/hcls-fhir-rdf/spec/rdf-SBS.html

tony: I'd rather review the new document.

eric: But the SBS doc drives the generation of the Turtle templates

sajjad: Are we driving toward a doc where an RDF expert can take the FHIR ont and start representing things in RDF?

eric: Guoqian and Harold are working on mapping structure def to templates for FHIR Turtle.
... We're working on that for a JAMIA paper, but when we're finished it may be helpful to this work.

sajjad: The other way around, a knowledge engineer takes FHIR RDF ont, and represents instances as RDF, rather than first knowing FHIR
... As a user, I want to use FHIR as just another ontology.

harold: That's dangerous territory, because the FHIR ont only talks about medical records -- information models, not what they are about.
... You may went to look at SNOMED and LOINC instead.

sajjad: If I want to represent clinical information model . . .

tony: We already have a model, it's the FHIR model. We're just representing it in RDF.
... The template notation is a way to show how to write it.

harold: And ShEx is to the RDF as XML Schema is to XML.

eric: Two representations of the schema: the RDF template, and ShEx. There's also the ont, which does not map 1-1 to the structure def.
... But there's a difference between an ont for an info model than one for OBO.
... A little of that nuance can be captured in the ontology.html doc.
... Do we have anything to allow an RDF geek to quickly grok the FHIR RDF model?
... Might need another doc for that. Or maybe the ontology.html doc could also fill that role.

dbooth: want a concrete plan forward

sajjad: to learn FHIR RDF, I can read the examples. Another way is to read the FHIR spec and check the schemas.
... As an RDF person, why not learn FHIR RDF?

lloyd: At some point you'll be able to focus on the FHIR RDF piece? But right now the RDF is not as prevalent as the other serializations.
... But there's much more the FHIR than the syntax. You need to learn the other aspects of FHIR also.
... We want RDF to be an equal sibling to XML or JSON.
... We don't have material written yet to guide the RDF reader.

dbooth: Proposals forward I've heard: 1. review eric's latest draft http://w3c.github.io/hcls-fhir-rdf/spec/rdf.html 2. review the previous draft http://w3c.github.io/hcls-fhir-rdf/spec/rdf-SBS.html

tony: We've already reviewed the SBS doc -- that was the side-by-side doc.

eric: things are missing

dbooth: How do we identify the missing things?

eric: They arise in doing the ShEx for them. Harold could list things that are not covered.

<hsolbrig> Apologies - gotta go. Will try to assemble a list as per eric

eric: Each section with an instance should have the logical model (table) from FHIR, and some ShEx, and then also instances.

dbooth: Our work plan is to look up the logical model and insert the corresponding ShEx into the SBS document?

eric: maybe we would keep the SBS document, but first it will be used to capture our model decisions.

dbooth: Take out the OWL from the SBS doc?

eric: Yes. It is already in the ontology.html doc.

ADJOURNED

========== 5pm Call =============

dbooth: ontology.html target audience will be OWL users who want to use the downloadable OWL FHIR ontology.

ontology.html document

eric: ontology.html should say how to use the ont.

tony: that will also have the mapping from FHIR model to OWL?

eric: That's what the mayo JAMIA paper will describe. want to help with it?

Reviewing the Ginormous ShEx

<ericP> ginourmous ShEx

<ericP> Constellations document

<ericP> -> https://github.com/w3c/hcls-fhir-rdf/blob/gh-pages/data/definitions.shex#L52 <Age>

grahame: Single file for ShEx, or how would we modularize it?

eric: Can do it either way. Might be useful both ways.
... In schema I generate modular schemas, and a combined schema, but some tools cannot cope with circular references with modular schemas.
... line 63 AllergyIntolerance is a type name?
... Sort of. It's a shape.

grahame: Why does it say it is an instance of itself?

eric: That's the difference between a shape and a type.

[[

<AllergyIntolerance> {

a [fhir:AllergyIntolerance]?,

]]

eric: <AllergyIntolerance> is a shape; fhir:AllergyIntolerance is a type.

grahame: Is that saying that it might have an rdf:type stated in the data?

eric: yes.

grahame: Why is it optional? Why not mandatory?

eric: It is only mandatory for the root.

dbooth: I think we decided to make it mandatory if it is a root or if it is an embedded resource.

grahame: Line 65 ':AllergyIntolerance @<DomainResource>*,' ?

harold: That's inheritance from DomainResource, but it's not right.

grahame: What's the difference between ":" and "fhir:"

eric: They should all say "fhir:"
... line 64 is a small set of possible values in square brackets

Topic Ginormous ShEx - fhir:Age

grahame: Age is a profile technically

dbooth: If Age is a profile, then the type should be the original type -- not a new type.

grahame: It should be element.id also
... If you generate an AllergyIntolerance you'll see issues.

harold: When we're generating from the meta we've been using the snapshot. We're using the JSON rep of the element definition.

grahame: Every element has a base that tells you what it comes from, with a base path.
... Use that.

harold: we had been liberally minting new predicates even if there was no difference from the root.

grahame: I'll need to build the ShEx in the build process. Are you using java?

harold: python. how should we dovetail with yours?

grahame: I'll put in a template Java class passing in the structure definitions, and we can both work on it there.
... Once we can generate the ShEx i'll put it into the build.

harold: I'm looking at Medication Structures (sec 1) in the constellations doc: https://www.w3.org/2016/FHIR-tutorial/Constellations

FHIR exmaple: http://hl7-fhir.github.io/medicationexample2.ttl.html

grahame: Look at this example;

http://hl7-fhir.github.io/medicationorderexample1.ttl.html

grahame: see http://hl7-fhir.github.io/medicationorder.profile.json.html
... notice "path": "MedicationOrder.id",
... and

"base": {

"path": "Resource.id",

<scribe> ACTION: Grahame to create a stub class for the ShEx generation, then harold will work on it with Grahame [recorded in http://www.w3.org/2016/04/12-hcls-minutes.html#action01]

<trackbot> Error finding 'Grahame'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.

dbooth: What work process?

harold: We will update the build process and review pieces as we go.

dbooth: Let us know what parts you have updated, so that we can review them.

grahame: We should then validate the ShEx against the Turtle examples.

harold: I'll work through the examples in the SBS document http://w3c.github.io/hcls-fhir-rdf/spec/rdf-SBS and do the ShEx for them
... And I'll be guided by Grahame's RDF
... And I'll also work with Eric to update the Constellations doc to align

dbooth: Ok, so let us know which example you convert and are ready to review.

ADJOURNED

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

12 Apr 2016

Attendees

Contents

Planning for Montreal HL7 Meeting Tue May 10, 11:00am-12:30pm

RDF page

========== 5pm Call =============

ontology.html document

Reviewing the Ginormous ShEx

Summary of Action Items

Summary of Resolutions

Scribe.perl diagnostic output