See also: IRC log
dbooth: Scheduled Tue Q2 in ITS group
http://www.hl7.org/events/wgm052016/
dbooth: Ken Lord, MDMI specialist, wants to present his work in order to align with RDF work.
tony: MDMI is an alternative to ShEx and XSLT
dbooth: Ken and I have drafted 4
use cases. We're organizing something about mapping for
Montreal.
... It would be good to encourage synergy.
lloyd: There's a new FHIR
resource being defined to do what MDMI does, but not MDMI. They
evaluated MDMI and decided that somethign FHIR specific would
be better.
... If you look at StructureMap in the spec you'll see.
dbooth: Anything scheduled on it?
lloyd: Nothing official. Don't know about BOF.
tony: Why FHIR specific?
... Maybe Ken could present the rationale for being FHIR
specific.
dbooth: Who will be in Montreal?
lloyd: Trying to get Grahame to come. I won't be at the ITS section.
harold: Will do my best to be there.
tony: I won't be there.
rob: I'll be there. Will bring teleconference gear.
dbooth: Status update, and review work products.
<Marc_Twagirumukiza> I will try to follow the TCon (thanks to Rob for organizing)
tony: And timelines. DSTU3 in summer?
lloyd: Voting will be in Aug;
reconciliation in Sept.
... Deadline mid-July.
... Question about whether we'll be able to call it DSTU.
tony: Still have some pieces we have not yet explored.
dbooth: RDF meeting in Montreal Will be 11-12:30 Tuesday.
sajjad: Would be good to send date/time in email.
https://hl7-fhir.github.io/allergyintolerance-example.ttl.html
dbooth: It uses fhir:nodeRole fhir:treeRoot; to indicate the root node.
tony: Looks close to what we discussed.
dbooth: Not sure that we've fully closed the question of what the root element URI should be or whether it should be a URI.
https://github.com/w3c/hcls-fhir-rdf/issues/18
tony: We could say http://example.org/ as the base URI
https://github.com/w3c/hcls-fhir-rdf/issues/3
dbooth: resource may not have a URI if it is being submitted.
harold: example.org should also be in the references
tony: It's in the FHIR spec
https://hl7-fhir.github.io/http.html
dbooth: I see: VERB
[base]/[type]/[id] {?_format=[mime-type]}
... I'll close issue 3 then :)
http://w3c.github.io/hcls-fhir-rdf/spec/rdf.html
eric: Changed to use the same
source as hl7 github page. That will make it easier to push
changes.
... If we focus on the RDF for resources (not structures or
valuesets) then we may be able to meet the July deadline.
... One view is that we can translate resources like
Observations, giving a default mapping and exceptions.
... Should we document that mapping or just hand it to
people?
... We'd like to be able to map an arbitrary resource with
confidence.
tony: I'm worried about the number of resources we need to get into OWL.
eric: The code generates a lot of the OWL. What else do we need to add to get more value added from the RDF?
tony: maybe we could do manual
transformation for valuesets and structure defs.
... Transformation from VS def in FHIR to the way it is
expressed in OWL.
lloyd: We will add to the build process a generation that takes every VS and every structure def and output the OWL for them.
tony: Need to define what we want it to be.
dbooth: I think of the ont page
as being for users of the ont, to tell them how to use
it.
... But it could also explain how the ont is generated.
lloyd: that should be in a secondary page.
eric: Harold, Josh and I are
writing a JAMIA paper that explains we generated OWL.
... FHIR does not publish how you map from a structure def to
the XML schema.
dbooth: FHIR defines that mapping by example
eric: If people want to help on the JAMIA paper they're invited. That paper might describe what tony wants -- defining the OWL corresponding to the structure def.
dbooth: Could we use the JAMIA paper as a starting point for a page that describes how the FHIR structure def is mapped to OWL?
eric: Yes
dbooth: Best path forward? Look
at examples and state the general rule?
... Maybe we should get Grahame to start generating some of the
OWL, to see how it looks
... Michael van der Zel also volunteered to generate the
OWL
... Propose starting with AllergyIntolerance, since we have
Tony's example of that
Tony: cannot make it
can make it: eric, harold, rob
dbooth: I can also make the 5pm call.
ADJOURNED
RESOLUTION: Close issue 3 as discussed at http://www.w3.org/2016/04/05-hcls-irc#T15-40-16
trackbot, start meeting
<trackbot> Meeting: Semantic Web Health Care and Life Sciences Interest Group Teleconference
<trackbot> Date: 05 April 2016
<hsolbrig> Hi folks. Lost track of time - just joining in
<hsolbrig> That said, the meeting number doesn't seem to work
[[
Tuesdays, 5:00pm Eastern US (Boston) time zone
Webex: https://mit.webex.com/mit/j.php?MTID=m5cd1bd8bb36825b9c4b369fd664bbb62
Teleconference: +1-617-324-0000 Access code: 645 777 110
Meeting password: 4257 ("HCLS")
]]
<hsolbrig> ah. I used 644 184 695
harold: Looking for a VS server for testing.
lloyd: Ping grahame.
harold: Trying fhir:Observation
in schema.org
... Not sure of the value. Original focus was shopping, using
semantic markup of html.
dbooth: Discussion of schema.org in the FHIR group?
lloyd: A long time ago. Don't remember where we landed.
harold: schema.org is branching
out from primary application (html). For stuff that got
generated, you could use RDF markup for provenance.
... But the more interested case would be moving unstructured
or semi-structured data into FHIR -- blog or device data.
... Promised the folks at biocaddie.org to try generating a
schema.org extension.
... You can tie the extensions into the schema.org core, but
because it is so focused, I think we'd want to be cautious on
doing that.
... We're doing our own. Classes, properties, etc.
... They have lots of nice tools for generating
documentation.
... What we want to put into schema.org is coarser than what we
want in FHIR.
... Or breaking datatypes into components does not seem
worthwhile for marking up HTML.
... In contrast with what we're doing with RDF, we want to
expose a subset of the differentials.
... SpecificObservation specializes Observation . . . I only
want to specify the differential.
... But i'm finding in looking at some of the specialized
Observations that, even though Grahame did not change the
types, they were still being included in the differential.
lloyd: Instance data or profiles?
harold: VitalSigns profile
specializes Observation profile. In some of the bits in the
differential section, they have types that are the same as in
the Observation profile.
... In the JSON
lloyd: Looking at the differential section
harold: I think the intent is the change the cardinality and fix code, but the type is there too.
lloyd: It's broken. It's not
slicing, but I would expect it to be slicing by system.
... Please submit a change request.
https://github.com/w3c/hcls-fhir-rdf/issues/9
https://www.w3.org/2014/12/02-hcls-minutes.html#item03
dbooth: I think we need to have the RDF express what the FHIR says, and before doing inference you need to do another step to check for unknown modifier extensions, and not try to inference over those sections.
lloyd: You need to first filter
out unknown mod ext.
... But how do you make the ones that you do recognize work in
the ont?
... You'll need to override some of the triples.
... My leaning for the first release is to say that reasoning
in the presence of known or unknown mod ext is challenging, and
guidance will be provide when we get more experience. i.e.,
punt. :)
... I'd like to try tackling from a real world perspective. My
guess is that there will be a few patterns that show up.
... Maybe make recs once we learn.
dbooth: Don't want this to hold
up getting a first release.
... This means including the modifier extension statements in
the FHIR RDF, and leave them up to the user.
lloyd: Should say that you cannot safely reason in the presence of mod exts.
dbooth: sounds good to me. I don't think we can do more at this point. it would mostly affect the descriptive pages rdf.html and ontology.html
RESOLUTION: Include caveat about mod exts saying that you cannot safely reason in the presence of them.
https://github.com/w3c/hcls-fhir-rdf/issues/11
https://www.w3.org/2015/04/22-hcls-minutes.html#item02
lloyd: Leaning to say it's not
the code, it's a profile. Because you'll need to take into
account other things, such as status=refuted, rule-out status,
entered-in-error, etc.
... Class would be better handled in the profiling space.
harold: In the instance space, do
we say that this is an instance of appendicitis.
... OWL inferred that people had integers, and Jane cannot be
both a person and a record describing something.
... You'd be saying that a data record is really an instance of
appendicitis, and that won't work.
... Wouldn't hurt to have an rdf:about in the profile space,
but it may be redundant.
... It's a portion of the record.
dbooth: what would this affect?
what would we do differently if the answer were one thing vs
another?
... i.e., do we really need to solve this, or can we leave it
to others?
... Seems to me that this question shows up when you try to
connect FHIR data to other ontologies like OBO that deal with
real world entities.
... But FHIR deals with records.
... What should we say?
lloyd: Say that we are representing records, and reasoners that want to reason about facts then additional logic may be need to extract the info that they need.
dbooth: Would be nice to show an example of how it can be done.
rob: Same question applies to the SNOMED CT ontology.
dbooth: Maybe a better example would be to show connection to OBO.
<hsolbrig> oops - hit hang up rather than mute...
harold: Include tony's examples?
lloyd: On a time-available basis.
RESOLUTION: Say that FHIR RDF is representing records, and reasoners that want to reason about facts may need to use additional logicto extract the info that they need.
RESOLUTION: Include tony's bee sting inference examples as time permits
https://github.com/w3c/hcls-fhir-rdf/issues/16
lloyd: Should treat all URIs the
same, not only Reference.reference
... We cannot use the URI type, because we are not doing things
in a pure URI way. Relative references are not evaluated in a
standard RFC 3986 way.
... Rel ref is resolved according the URL of the resource
containing the ref -- not the node that the ref is on.
... Also when we have a local ref (hash) pointing to an ID,
that is not resolved across the entire instance, it is only
within the containing resource.
... So we're better off treating them as strings in RDF.
dbooth: RDF itself has no
relative referenced. All URIs are absolute in RDF.
... So it would be possible, but messy to capture them, because
for round tripping we would also need a way to be able to turn
them back into relative URIs.
harold: Making them strings would
concern me, because URIs are so fundamental to RDF and LInked
Data.
... Could we deal with the funky FHIR rules and make it
work?
lloyd: No way to make it
work.
... E.g., patient with a contained org, #1. And a prescription
with a contained med, also #1.
... Then I query that causes them both to come back inside a
bundle. They are both digitally signed, so i cannot change the
URIs. No way to make the scoping document wide instead of
resource wide.
... Fundamental problem is that XML always thinks about the
document as immutable. But we're generating documents and
combining them.
dbooth: We'd be giving up value by not using URIs, because they're so fundamental to RDF.
lloyd: If we want to have a URI that points to something, we could have a property.
harold: It's not 'pointsTo' that matters, but the identity issue: is this the same as that, so that we can like the RDF together, regardless of whether it dereferences.
lloyd: May want an additional property that is the full URI of the reference. Take the ref and if it's relative make it absolute.
dbooth: i.e., leave the relative
URI but also have an absolute URI.
... Sounds reasonable, but want to seen an example.
This is scribe.perl Revision: 1.144 of Date: 2015/11/17 08:39:34 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) No ScribeNick specified. Guessing ScribeNick: dbooth Inferring Scribes: dbooth WARNING: Replacing previous Present list. (Old list: David_Booth, Lloyd_McKenzie, Harold, Solbrig, Marc_T, Rob_Hausam, Sajjad_Hussain, Thomas_Lukasik, Tony_Mallia) Use 'Present+ ... ' if you meant to add people without replacing the list, such as: <dbooth> Present+ 5pm_call:_David_Booth, Harold_Solbrig, Lloyd_McKenzie, Rob_Hausam Present: 5pm_call:_David_Booth Harold_Solbrig Lloyd_McKenzie Rob_Hausam Agenda: http://wiki.hl7.org/index.php?title=ITS_RDF_ConCall_Agenda Found Date: 05 Apr 2016 Guessing minutes URL: http://www.w3.org/2016/04/05-hcls-minutes.html People with action items:[End of scribe.perl diagnostic output]