Semantic Web Health Care and Life Sciences Interest Group Teleconference -- 29 Mar 2016

Draft of draft FHIR RDF page on github

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

* Marc_Twagirumukiza (~Marc_Twagirumukiza@public.cloak) has joined #hcls

Eric: Tony looked at FHIR XML and modeled FHIR RDF using Protege and turtle. Side-by-side doc was a to-do list, but not oriented to the specs on the HL7 site.
... Also overhauling: now discussing logical model instead of XML.
... For reference we can look at the JSON represenation of Observation

<ericP> http://hl7-fhir.github.io/observation#resource

eric: The Structure tab at that page shows the logical model of fhir:Observation
... It defines the official FHIR data model for Observation
... The XML and JSON tabs show how it looks in XML and JSON, defined in terms of examples.
... Example JSON
... It's a template that you can almost copy and paste for a working example.
... If you click the XML tab it gives you more info, such as the cardinality

marc: The descriptions are automatically generated?

eric: They're auto generated from a publication process that starts with spreadsheets but ends with profiles. The text is in the spreadsheets.

marc: Is the JSON auto generated from the XML?

eric: No, from the structure definition.

tony: I think the original is in eclipse ecore
... you can also get the StructureDefinition written in XML, if you click where it says "Alternate definitions" below the table: http://hl7-fhir.github.io/observation.profile.xml.html

eric: I've been working on an RDF version that follows this motif

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

eric: At the bottom of this doc in the "stuff to move elsewhere" is a Turtle-ish example corresponding to the XML and JSON tabs
... I am also doing a ShEx version, which will be machine processable.

tony: Which is normative?

dbooth: Normally the prose version would be normative, because it can express any constraints, whereas formalisms are limited. If the formalism disagrees then it indicates a bug, which we need to fix.

eric: If we follow the FHIR approach, the profiles will be normative, and then the ShEx, OWL and RDF templates are all slaves to that, and if they are wrong then it's a bug to be fixed.

tony: Ontology is comparative to XML schema.

eric: ShEx is more comparable
... The XML model is defined in part in XML Schema and in part in Schematron.

tony: If you generate all three from the source then fine.

eric: There was a FHIR freeze deadline at 5pm today. I probably won't make that deadline, but I might be able to get it onto the HL7 site.

tony: What about the namespace prefix? an empty one?

dbooth: Safer to use the fhir: prefix. People will imitate and copy what the see.

eric: Plan is to change everything to be compared to the logical model, and provide two renderings: faux-Turtle and ShEx

dbooth: Coloring the faux-Turtle will help also
... Great progress!
... Next steps and help needed?

eric: Want to spend another week with it, then see what others think.
... Goal is to break into 3 things: 1. fhir.org RDF page (like the XML and JSON pages), showing how extensions work, leaf values, etc. Not super formal, but must be intuitive like the XML and JSON pages.
... But some pieces will go in other places, such as the Observations description.
... Once we get that in, grahame is very quick about getting it deployed into the build process.
... There is also a bunch of stuff that talks about ontologies. Trying to figure out what to do with that. Some should go into ontology.html document, but some has already been reworded.
... Need to aim for SPARQL heads in the faux Turtle.
... Purple marker along the edge in the doc means it is also in the ontology.html doc.
... Examples should probably be in both; prose not.

<ericP> http://w3c.github.io/hcls-fhir-rdf/spec/rdf-SBS.html

eric: In the ontology.html doc, what's not there is the structure of the RDF as defined by what comes from the structure defs.
... That's not needed there.

tony: Should we do the structure def to OWL mapping?

dbooth: sounds like a good idea
... And it will be put into code in the build process too

Funding for W3C HCLS interest group

Eric: My funding comes from W3C to work on this.
... HCLS interest group does not have enough paying members to pay me to work on it.
... Two choices: Reduce expectations by turning it into a Community Group, which means that we cannot publish the same way.
... Or get another full member, such as Accenture
... i would rather scale up than down. Lots of opportunity to collaborate on data shapes work, schema.org work and other.
... As in Interest Group, there's a tech plenary in Lisbon in September, we could meet there.

dbooth: Should people try to get their organizations to become members? If so how?

eric: yes. Affiliate membership is $7800, Full membership is ~$70k
... Are there EMRs, big clinical systems, or someone who serves as code-in-the-middle like Accenture, it would be good to pursuade them.

marc: Tricky. I had the same discussion with Agfa. They used to be members, but hard to convince.
... Can we draft a grant request, maybe $100k to NIH?

eric: NIH is tricky to get money out of. They have less jurisdiction about how they spend their money. Might be easier to go to a granting foundation. But the other problem is when renewal time we'll need support from at least 20 W3C members to renew the group.

lloyd: Not clear on what the relationship within W3C compared with HL7 and fhir.org . Work on this spec should be part of the FHIR spec. Cooperation with other W3C and affiliated efforts should be ongoing.
... Not clear on the scope.

eric: This started within HCLS group as COI (clinical observation interop)

dbooth: The FHIR RDF work began as a joint W3C-HL7 effort.

rob: Cecil has interest at some level. Could explore that connection again.

thomas: given the lack of clarity on W3C HL7 roles, would it be an impediment to funding?

dbooth: Next steps?

<scribe> ACTION: Eric to talk to Cecil Lynch about possibly Accenture becoming a full member [recorded in http://www.w3.org/2016/03/29-hcls-minutes.html#action01]

<trackbot> Created ACTION-52 - Talk to cecil lynch about possibly accenture becoming a full member [on Eric Prud'hommeaux - due 2016-04-05].

If we cannot get enough funding, then we'll have to change to a Community Group.

5pm call

dbooth: Who can make it?
... eric, rob, lloyd, dbooth

Lisbon face-to-face meeting in September?

eric: Could meet with the schema.org group. Their clinical part showed up again recently.
... Could also meet with the Data Shapes group and the Best Practices on the Web group

dbooth: International travel is harder. I would not be able to attend.
... Would anyone be able to make it? (Nobody but eric able to make it.)

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

Faux Turtle

ttp: //w3c.github.io/hcls-fhir-rdf/spec/rdf.html at the bottom is an example of faux-turtle

dbooth: Suggest putting the extra rdf:type arcs in comments: :Obervation.identifier : [ a :Identifier ];

grahame: I'll draft somethign next week that looks like the XML and JSON

eric: Suggest ':myAllergyIntolerance fhir:nodeRole fhir:treeRoot' on root node
... because i got pushback on using a boolean predicate

grahame: fine

dbooth: ok with me

RDF Page

grahame: what to do with the rdf.html page in the next few hours?
... I could also generate shex in the doc. That would allow mini-connectathon in Montreal, to allow validation against the ShEx.

dbooth: That woudl be an excellent forcing function.

grahame: how far through page http://w3c.github.io/hcls-fhir-rdf/spec/rdf.html should i try to put into the hl7 rdf.html page?

eric: until Resource Root
... it would be nice if the JSON, XML and turtle all used the same bits of an observation.

dbooth: could then try round tripping it

Identifying the root node

eric: Suggest ':myAllergyIntolerance fhir:nodeRole fhir:treeRoot' on root node

dbooth: Any objections to changing to that? eric got pushback on using a boolean.

(no objections)

Date literals

https://github.com/w3c/hcls-fhir-rdf/issues/23

Prior Discussion:

https://www.w3.org/2016/03/01-hcls-minutes.html#item06

dbooth: Are coercions between xsd:gYear xsd:gYearMonth xsd:date xsd:dateTime a part of standards, or just commonly implemented in SPARQL engines?

eric: Standard casting in SPARQL, which delegates the spec to XPath

dbooth: Therefore there would be convenience in using those datatypes explicitly in FHIR RDF.
... But that means that the FHIR XML/JSON --> FHIR RDF translator must check each FHIR date literal to see which xsd type it matches.

grahame: this comes up with numerics also

eric: but printf will handle those. dates are the only ones that are unions.
... At present the converter to FHIR RDF does not handle this date stuff yet.

dbooth: Are we agreed that FHIR RDF should generate the specific 4 xsd date types? Any objections?

(No objections)

AGREED: FHIR RDF should generate the specific 4 xsd date types

AGREED: Use xs: prefix instead of xsd:

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

29 Mar 2016

Attendees

Contents