Semantic Web Health Care and Life Sciences Interest Group Teleconference

23 Dec 2014

See also: IRC log


Cecil_Lynch, Claude_Nanjo, David_Booth, EricP, GJiang, Ingeborg, John_Mattison, Joshua_Phillips, Lloyd_McKenzie, Patricia, Paul_Knapp, Peter_Hendler, Scott_Marshall, Tony_Mallia, Rob_Hausam
David Booth and Paul Knapp


<trackbot> Date: 23 December 2014

Talky is here: https://talky.io/hcls

We will try using https://talky.io/hcls for audio

Talky is here: https://talky.io/hcls

Talky is here: https://talky.io/hcls

<drjava> Drjava is peter hendler

C-CDA as ORIM RDF (Eric Prud'hommeaux)


Eric: Charlie and I created the ORIM -- translation of RIM to OWL.
... a few years ago.

<ericP> http://www.w3.org/2012/12/O-RIM/protege-tutorial/simple.omn

<ericP> http://www.omg.org/spec/ODM/1.1/PDF/

Eric: IThe translation into OWL was done via the above linked document.
... It assumes that relation names in UML are not unique
... so it prefixes every relation name with the name of the class it came from.
... So a predicate for a car with a registration would be car.registration.registrationDocument
... This makes predicate names very long, but they work.
... Lloyd did this for doing inference on the RIM model.
... It's an enormous ontology.
... My intention was that it be useful to turn CCDA XML into RIM, then turn RIM into RDF.

<Claude> Test

Eric: so this document: http://www.w3.org/2013/C-CDA/IJ.xml was my attempt to not be creative, but to produce the same RDF that anyone would.
... CCDA example came from Josh Mandel. In principle this xslt should work with any CCDA.

Joshua: I'm specifically interested in including genomic and phenotypic data.
... But basically want a concrete RDF representation that we can query using SPARQL.

<Claude> To David: Could you please paste the link again. I just installed an IRC client.

The page eric is discussing: http://www.w3.org/2013/C-CDA/IJ.xml

Eric: The main arcs of connectivity in RIM are ACT relations.
... Using vocab from HL7 indicating severity, etc.
... In principle those arcs make the RDF graph.
... One way to make this more like what RDF people would use would be to turn these ACT relationships into RDF arcs, which means throwing out some information.
... This would make a lean and RDF-friendly way to represent clinical data.
... Then we could do RDF tricks from there.
... Does it make sense to collapse the ACT relatinoships into predicates?

Lloyd: No, because there are a lot more predicates attached to them.
... Need to go through the TYPE codes to see which ones are identified.
... When those other things are there, they are important.
... My leaning is to have that detail there always, but have the shortcut there as well when it can be done.

Cecil: What would be the use case for an RDF model that drops content?

Eric: If the majority of ACT relationships were simple TYPE codes, then one could write SPARQL queries that use those
... and when there was something else attached, then you would fall back to a more complex sparql query.

cecil: to be safe for clinical use, you can't ever drop information.
... Not safe to drop them ever.
... But if our use of this is FHIR then why not use FHIR where it is flattened?
... If you're going to model the RIM or CDA in RDF, you should not drop any of the model, because it would be dangerous to lose clinical content. Would risk clinically significant error.

Eric: Not planning on discarding info, but using simple arc when there is ONLY a type code.
... The other goal was to map this to FHIR RDF.
... Wrote a script to spit out XSLT and ShEx code that will round trip FHIR XML to FHIR RDF


<ericP> ttp://www.w3.org/2014/Talks/1209-shex-egp/#(24)

<ericP> ttp://www.w3.org/2014/Talks/1209-shex-egp/#(24)


<ericP> http://www.w3.org/2014/Talks/1209-shex-egp/#(45)

Eric: ShEX is good for validating RDF, but also (slide 45) good for translating


Eric: This is a complex example. What's shown at the top is turned into what's shown at the bottom.
... There are actions specified when the pattern on the left is matched.

David: When the pattern on the left is matched, the %GenR action is run?

Eric: Right.

<ericP> http://www.w3.org/2013/ShEx/FancyShExDemo?schemaURL=test/O-RIM-to-FHIR-Patient/schema.shex&dataURL=test/O-RIM-to-FHIR-Patient/pass-IJ.ttl

Scott: the leading zeros are missing from the dates.

Eric: yes, didn't fix that yet.

David: now looking at the above link, shown Schema (ShEX) on the left and Data (Turtle) on the right.

Eric: The left is the input schema, the right is the input instance data, and below is the validation result.
... when you click (at the bottom) "View GenR output as popup" then you see the output.
... FHIR instance data could be combined with schema.
... then you could conclude an allergy to bee stings
... Want to see if this would be good enough for use by the CDA work group at HL7.
... Wanted to see whether we could migrate between different information models and combine that with validation, and this seems to do it.
... Also the RDF Data Shapes working group is deciding what language to use, so comments on ShEx are invited.

David: It reminds me of yacc rules.

Paul: Is the intent of validation to determined a valid instance?

Eric: Yes, validate an instance at an RDF graph level.
... The semantic actions can do more business logic validation that the patterns cannot validate.

Paul: Efficiency?

Eric: Paper on Regular Bag Expressions looks at that.

<scribe> ACTION: Eric to supply link for paper on Regular Bag Expressions [recorded in http://www.w3.org/2014/12/23-hcls-minutes.html#action01]

<trackbot> Created ACTION-14 - Supply link for paper on regular bag expressions [on Eric Prud'hommeaux - due 2014-12-30].

John: Also interested in compelling complex examples

<ericP> http://researchers.lille.inria.fr/~staworko/papers/staworko-icdt15a.pdf

<scribe> ACTION: JohnMattison to provide example CCDA for trying with ShEx [recorded in http://www.w3.org/2014/12/23-hcls-minutes.html#action02]

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

Approve minutes


Minutes unanimously approved!


<ericP> (figuring tony's phone infrastructure failed to send the "all done")

Summary of Action Items

[NEW] ACTION: Eric to supply link for paper on Regular Bag Expressions [recorded in http://www.w3.org/2014/12/23-hcls-minutes.html#action01]
[NEW] ACTION: JohnMattison to provide example CCDA for trying with ShEx [recorded in http://www.w3.org/2014/12/23-hcls-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2014-12-23 17:20:38 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.140  of Date: 2014-11-06 18:16:30  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/the input/the input schema/
Succeeded: s/the output/the input instance data/
Succeeded: s/Maddison/Mattison/g
No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Default Present: DBooth, bpech, +1.410.624.aaaa, Joshua_Phillips, Ingeborg, knapp, +1.510.418.aabb, PeterHendler, +1.727.501.aacc, +31.62.427.aadd, Patricia, [IPcaller], mscottm, cecil, EricP, Tony, rhausam, JohnMaddison
Present: Cecil_Lynch Claude_Nanjo David_Booth EricP GJiang Ingeborg John_Mattison Joshua_Phillips Lloyd_McKenzie Patricia Paul_Knapp Peter_Hendler Scott_Marshall Tony_Mallia Rob_Hausam
Found Date: 23 Dec 2014
Guessing minutes URL: http://www.w3.org/2014/12/23-hcls-minutes.html
People with action items: eric johnmattison

[End of scribe.perl diagnostic output]