W3C

- DRAFT -

Semantic Web Health Care and Life Sciences Interest Group Teleconference

16 Dec 2014

See also: IRC log

Attendees

Present
David_Booth, Kerstin_Forsberg, Paul_Knapp, Tony_Mallia, Rob_Hausam, EricP, Ingeborg, Brian_Pech, Charlie_Mead, Cecil_Lynch, Scott_Marshall, Claude_Nanjo, Vipul_Kashyap
Regrets
Lloyd_McKenzie, Hans_Cool
Chair
David Booth and Paul Knapp
Scribe
ericP

Contents


<trackbot> Date: 16 December 2014

<Ingeborg> zakim aaee is ingeborg

<dbooth> zakim Brian_Peck is really bpech

charlie, [IPcaller] is claude

<dbooth> \

<scribe> scribenick: ericP

Approval of Minutes

<dbooth> http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20141209

RESOLUTION: approve http://wiki.hl7.org/index.php?title=ITS_RDF_Concall_Minutes_20141209 as a record of the dec 9 call

Action Item Review

<dbooth> ACTION: Tony to find out more details about how iCat handles ICD-11 ont and report back [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action01] -- PENDING

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

<dbooth> ACTION: Guoqian to figure out whether he can share URI conventions for ICD-11 [recorded in http://www.w3.org/2014/11/25-hcls-minutes.html#action07] -- PENDING

<trackbot> Created ACTION-12 - Figure out whether he can share uri conventions for icd-11 [recorded in http://www.w3.org/2014/11/25-hcls-minutes.html#action07] -- pending [on Guoqian Jiang - due 2014-12-23].

<dbooth> ACTION: Kerstin and Ingeborg to prepare a status and future state ideas for PhUSE-FDA work [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action05] -- PENDING

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

dbooth: i have Frederick @@1 to speak on @@2

<dbooth> ACTION: Tony and all to decide on a wiki for Term Info work [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action09] -- DONE

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

-> https://www.w3.org/wiki/HCLS/ClinicalObservationsInteroperability/C-CDA C-CDA -> RDF page

<dbooth> Rob: I did that action

<Claude> https://global.gotomeeting.com/join/157514853

<Claude> Please join the gotomeeting

<dbooth> ACTION: Tony and Rob to report their plan on High-level concept mapping to RDF work [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action08] -- DONE

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

<Vipul> Thanks Claude

<Claude> https://global.gotomeeting.com/join/157514853

<dbooth> ACTION: Eric and Joshua to report on C-CDA RDF representations work plan [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action07]

<trackbot> Created ACTION-13 - And joshua to report on c-cda rdf representations work plan [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action07] [on Eric Prud'hommeaux - due 2014-12-23].

<dbooth> DONE

<dbooth> Tony: Will send email to the list.

<RHausam> http://wiki.hl7.org/index.php?title=TermInfo_FHIR_Ontology_HL7_ITS_RDF_W3C_HCLS_COI

Cecil Lynch's FHIR Ontology review

-> http://wiki.hl7.org/index.php?title=TermInfo_FHIR_Ontology_HL7_ITS_RDF_W3C_HCLS_COI

Cecil: followed lloyd's point that we need an automated, repeatable way to do this
... i've used TBC to import XML Schema into RDFS and OWL

<dbooth> cecil: I imported XML Schema of FHIR into RDF/OWL using TopBraid Composer.

<Claude> To David: Could we reserve 5 min at the end of the meeting to discuss plans for Jan meeting in San Antonio?

<dbooth> claude, good idea

<RHausam> Sorry, Eric - must be dyslexic this morning. :)

Cecil: you can tell composer to import schema and optionally instance data
... i pointed all of the FHIR XSDs in a package

<dbooth> cecil: I pointed it to all of the XSDs and it imported all of them.

Cecil: you can have it prefix the object properties with a "has..."

<dbooth> ... Also imported all of the datatypes

Cecil: everything drills down to privimage types like [21090] Code

<dbooth> ... It's faithful to the XML Schemas, in an automated way.

Cecil: follows lloyd's goal to completely import the FHIR spec

<dbooth> ... The potential downside is that it brings in EVERYTHING, giving a complex model.

Cecil: downside: you get a complex model

<dbooth> ... We've built tooling around HL7, OWL interface to Jena API, and we want all the detail.

Cecil: at accenture, we've built tooling to use this with NLP, with links to Jena API

<dbooth> ... But for human mapping you might not want all of this complexity.

Cecil: you need to examine whether you need this level of detail
... [examining AdverseReaction]

charlie: i'm just staring with a company in paris which is the medication info provider for the french government

<dbooth> Charlie: I'm just starting with a Paris company, medication provider for French Gov, committed to semantic infra, using TBC.

<dbooth> ... Will be talking about importaing FHIR specs.

charlie: we have a call tomorrow with john hayden to talk about about the import from FHIR specs

<dbooth> ... Asked cecil if FHIR spec XSDs can be imported directly.

<dbooth> ... Also emailed Eric about his approach.

<dbooth> Eric: I'm driving this from XML Schema and Cecil is, so the only differences (aside from cosmetics) should be where I've customized toward the RDF-natural way to do things.

<dbooth> Cecil: I had it add "has" to predicates.

charlie: does ericP/Josh's instance map correspond to Cecil's schema?

ericP: in principle, apart from our customizations and coin flips in the parameters

Cecil: agreed

<dbooth> Tony: I looked at TBC and didn't use it, but got the specs on the XML->RDF transforms that they use.

Tony: i haven't seen a transform for a sequence yet
... sequence is interesting.

<dbooth> Tony: I suspect that we need to transform both the schema and the profile, to get more specific.

Tony: looking at the high level, we need to transform the schema and the profile.

<dbooth> ... If you define a link as being a reference, which is the explicit XML, what you want is that it's resolved to the patient.

Tony: if you define a link as a reference (which is the explicit XML), what you want is that it's resolved to the patient resource

Cecil: ...to the patient resource.

Tony: so the resource is the referent in the XML the way the XML defines the reference.
... so we need it to be reasoner-friendly

<dbooth> Cecil: I did generate an ont to do it that way, but you need to bring in the target resource.

<dbooth> Eric: I have a lab test for a patient, XML reference strcuter with attribute having a URL for that subject.

Cecil: i did that using TBC

Tony: the question is how do you get a reference that the reasoner can actually deail with
... this should be fine apart from the specific resource
... i believe those are in the profile

<dbooth> Tony: This looks fine other than the specific type of Resource -- how to restrict it to a patient reference.

<dbooth> cecil: Let me switch to a different ontology.

Cecil: looking at a 2.5 message spec
... looks a little different, but same idea
... sucked in the schema

<dbooth> cecil: this is an HL7 2.5 spec

Cecil: you have PID segemtn and patient content
... this was build to model the lab domain for biosurveillance

<dbooth> ... This was built to model the data for biosurveilllance

Cecil: so i have an OBX and instance data that fits this model.
... that's done with terminology bindings
... drilling down on the instance data, i see the properties of that LOINC code
... this is how you'd bind that instance data to the schema
... this is fully machine-processable
... in practice, we don't do with with inference 'cause it would take forever

<dbooth> cecil: once you have validated your ont, you can do instance data inference some other way.

<dbooth> eric: A difference from the way Tony versus Josh and I took, in one case we did mechanical transformation and we did not dig into the coding systems -- terminology layer.

<dbooth> ... But in Tony's we dug in, for obs coding system, snomed code, allergic reaction that is bee sting.

<dbooth> ... What we're looking at here is a combination of mechanical transformation of obs records, followed by extra rules that could be modeled as OWL equiv classes

<dbooth> cecil: Correct, and i have a tool in sourceforge

Cecil: i put a tool up in sourceforge that translate e.g. presence findings into a normal form
... it has lots of swtiches:
... .. taking a top-level SNOMED concept and turning things into instances or classes

<dbooth> ... tool lets you build vocabularies (with switches to include skos, etc.), set top level class with terms of instances, and it auto imports snomed, rxnorm, loinc from their native files.

Cecil: .. imports terminologies from their native files

<dbooth> ... You point it to whatever comes and it imports terminologies.

Tony: i use the SNOMED export to classes

<dbooth> ... This allows you to export as classes or as instances.

<dbooth> Tony: That's been a debate: the two ways to represent concepts (as classes vs instances).

Tony: this brings up the question of modeling classes or instances

<dbooth> cecil: It's a matter of how to model it to use it in a production system.

<dbooth> eric: Is there a clear winner about the better approach?

Cecil: yeah, question of how to model for your use case

<dbooth> cecil: Depends on the use case.

<dbooth> ... TB surveillance system used classes, because in that case there is a generator that writes rules for them, and we use backward chaining in JESS

Cecil: i've modeled as classes for the CDC's turberculosis to then backward-chain in jess
... everything in that system is classes

<dbooth> ... CDC has a set of OWL classes and we validate against medications the person is taking

Cecil: we validate against instance data for e.g. medication the patient was taking

<dbooth> ... But this one uses instances because it is building value sets where i use a top level concept and I want everything to be an instance: does that value appear in that field?

Cecil: this one is a little different because it's building value sets where i don't care about hierarchy

ericP: so if you're doing a leaf closure, you go straight to instances

dbooth: @@1

Cecil: the real question for the group in general is what are we aiming for?
... some target use of the ontologies?

<dbooth> david: Like the difference between instance data validation as the RDF Validation WG is working on, versus class level inference like OWL reasoners do.

Cecil: 'cause if you're using it for large-scale data processing, you won't use reasoners.

<dbooth> cecil: None of the OWL reasoners are robust enough for production processing.

Cecil: i'd wrap it up in a jar and add it as a SPOUT in SPARK

<dbooth> Tony: Could also use punning, so we have both instances and classes

Tony: we could be looking at puns where the instance in one is a value in another

Cecil: yep, the code value is the owl:individual

<dbooth> cecil: I did use punning, and that makes it a lot easier to navigate from a machine processing viewpoint.

Tony: so in an instance, you can reference either IRI

FHIR Ontology requirement

<dbooth> http://wiki.hl7.org/index.php?title=FHIR_Ontology_Requirements

dbooth: any comnents about FHIR ontolog reqs?

Vipul: these reqs are about syntax

<dbooth> Vipul: These requirements are primarily targeted at syntax. Should look at use cases and then look at requirements.

Vipul: we should go back to real-world use cases and then go to reqs
... that approach was successful in the COI group

<dbooth> 1. It must be possible to round-trip instance data from XML/JSON through RDF representation

<dbooth> Eric: We have one shot to impress people with how cool RDF is in the clinical domain, if we want to focus our efforts, we want to pick something that is achievable and differentiates the value proposition that shows benefits of using RDF over XML or JSON, etc.

<Vipul> +1

dbooth: 1 seems like MUST

+1

<Tony> +1

ericP: is this met by round-tripping just to XML, counting on the compiles XML-JSON mapping

Vipul: can it be lossless?

dbooth: whether it's lossless or not has to be viewed through the lense of what is semantically significant according to the spec

Tony: the translated XML instance data has to be parsable

<RHausam> +1

Tony: e.g. TBC sequences doesn't parse

<dbooth> 2. We want to be able to represent instances as RDF and Profiles as OWL/RDFS

+1

<dbooth> +1

<dbooth> 3. Syntax needs to be "safe" when dealing with modifier extensions

+1

<dbooth> QUESTION: What is meant by "syntax"? RDF has several serializations. What syntax do you mean?

<dbooth> Eric: You can have extensions, so before interpreting results you need to check whether there are modifying extensions.

<inserted> Screen shots from Cecil's FHIR ontology discussion: http://lists.w3.org/Archives/Public/www-archive/2014Dec/0002.html

January HL7

<Vipul> Does anyone know the FHIR Link for "Version 2" - on Claims and Adminstrative Processes?

<Vipul> Can someone please post this here?

<Vipul> Thanks!

<dbooth> Paul: We could do Q1 or Q2 Monday for RDF subgroup.

<dbooth> Tony: Remote access?

<dbooth> Tony: We can try, with skype, but HL7 doesn't provide it.

<dbooth> Rob: I have audio hardware I can bring.

<dbooth> David: Who will be coming to San Antonio?

<dbooth> Answers: Claude, Paul, Cecil, Charlie, David

<dbooth> ... Rob

<dbooth> ... Brian

<dbooth> Tony: Out Dec 23

<dbooth> Charlie: Out Dec 23 and 30

<dbooth> Eric: I won't be able to attend the San Antonio, but I would love for people to be able to run the FHIR transformations that either Cecil or I have done.

<dbooth> ADJOURNED

-> https://github.com/jmandel/fhir-rdf/blob/master/generic/tests/other-nonmon.xml from

-> https://github.com/jmandel/fhir-rdf/blob/master/generic/tests/other-nonmon.xml to

-> https://github.com/jmandel/fhir-rdf/blob/master/generic/tests/other-nonmon.ttl#L15 to

Summary of Action Items

[NEW] ACTION: Eric and Joshua to report on C-CDA RDF representations work plan [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action07]
 
[PENDING] ACTION: Guoqian to figure out whether he can share URI conventions for ICD-11 [recorded in http://www.w3.org/2014/11/25-hcls-minutes.html#action07]
[PENDING] ACTION: Kerstin and Ingeborg to prepare a status and future state ideas for PhUSE-FDA work [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action05]
[PENDING] ACTION: Tony to find out more details about how iCat handles ICD-11 ont and report back [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action01]
 
[DONE] ACTION: Tony and all to decide on a wiki for Term Info work [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action09]
[DONE] ACTION: Tony and Rob to report their plan on High-level concept mapping to RDF work [recorded in http://www.w3.org/2014/11/18-hcls-minutes.html#action08]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2014/12/16 20:15:01 $

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/-< http/-> http/
Succeeded: s/patient/patient resource/
Succeeded: i/Topic:/Screen shots from Cecil's FHIR ontology discussion: http://lists.w3.org/Archives/Public/www-archive/2014Dec/0002.html
Succeeded: s/Topic: FHIR Ontology review with Cecil Lynch//
Succeeded: s/topic: FHIR requirements//
Found ScribeNick: ericP
Inferring Scribes: ericP
Default Present: +1.604.250.aaaa, +1.978.794.aabb, DBooth, Kerstin_Forsberg, +1.801.949.aacc, +1.678.999.aadd, +1.202.528.aaee, knapp, Tony, rhausam, ericP, Ingeborg, bpech, +33.1.73.71.aaff, +1.415.537.aagg, Charlie, cecil, +31.88.445.aahh, mscottm, [IPcaller], +1.617.498.aaii
Present: David_Booth Kerstin_Forsberg Paul_Knapp Tony_Mallia Rob_Hausam EricP Ingeborg Brian_Pech Charlie_Mead Cecil_Lynch Scott_Marshall Claude_Nanjo Vipul_Kashyap
Regrets: Lloyd_McKenzie Hans_Cool
Found Date: 16 Dec 2014
Guessing minutes URL: http://www.w3.org/2014/12/16-hcls-minutes.html
People with action items: eric guoqian ingeborg joshua kerstin tony

[End of scribe.perl diagnostic output]