harold: Two items getting in the
way. 1. I need to put type arcs on all of the code elements --
the ones that have an implicit system.
... My recollection: during discussion in dev days meeting, I
thought we decided not to do it in the spec, but Grahame felt
fairly certain we should put them in.
... 2. Get the OWL representation of the FHIR code systems
published.
... 3. One other thing grahame mentioned: we need to get
valueset information into fhir.ttl
david: Give a webinar on the I2B2 work?
harold: sure.
Tentatively scheduled as yosemite webinar 2pm Eastern on Apr 26.
eric: What about slicing?
harold: we haven't tried representing extensions as first-class elements, but there's huge value in doing it.
harold: Right now, when we get into extensions we go from URIs to fhir extension URI and code.
eric: Kind of reification.
harold: I suspect we should not
touch what we have now, but add additional code -- turn the URI
into an RDF node.
... Need to be able to remove it going in the other direction
(from RDF).
... Also fhir.ttl is a closed package -- it identifies every
predicate. If we start de-reifying extensions, then the ext def
itself needs a mechanism to pull them into our
reasoners/converters, bc we'll need to understand what the
predicate is.
david: Seems to me that we need some rules, to produce the de-reified from the reified, and vice versa.
harold: If you keep both the
reified and non-reified, what to do if they are inconsistent?
One says 2 the other says 3?
... Check it in shex?
eric: I don't think so. We don't
have variable predicates.
... Could also compile a shex schema for each extension. That
would match the business process.
harold: This pattern is well used in the FHIR world.
david: Also use running code?
harold: yes, we could.
eric: You tell the tool what extensions you want, and it generates the shex for you
harold: Need something that traverses a profile and generates additional vocab elements, to go along with fhir.ttl
eric: Also need to make sure that it works on extensions on extensions on extensions
harold: We start getting into the
valueset area, where we currently have a gap. We do small ones,
but big ones ....
... When we get into profiles, they start making the valuesets
finite, and we'll need to enforce. And that brings us to
slicing.
eric: I've been looking at it
from the shex perspective, but I'm not confident that I'm
covering the use cases.
... I have running code on the shex side, but the slicing
extensibility seems to be everything to all people.
... I would love help from somebody who can grok that slicing
spec and the reasonable use cases, and help me understand what
they mean.
harold: One thing that's made FHIR easier to deal with is the rule that if there aren't test cases and examples thenn it doesn't exist. So I think you could push back on features that lack examples and test cases.
david: Check with grahame. Maybe he can point you to someone.
<scribe> ACTION: Eric to ask grahame for pointer to test cases for slicing
<trackbot> Sorry, but no Tracker is associated with this channel.
eric: I need to be able to tell the difference between restriction and extension.
harold: In I2B2 we don't want to
talk about the obs whose code is hematicrit, we want to talk
about the hematocrit result.
... In the slicing core, they say each component code names the
thing. Where does that fit in?
eric: There are discriminators.
If we could communicate them to a shex engine, we could trim
some branches.
... The members of a slice are treated as predicates right?
harold: He's say the obs code is
the identity of the obs.
... In I2B2 we have a rule to turn it into the hematocrit
value.
... The rule generates a new triple.
... In profile, are we interested in doing what Stan discusses
in CIMI, turning into a hematocrit class.
david: Seems like a lot of this boils down to the need for rules language.
eric: what about your experience with shexmap for rules?
david: It can be used for rules, but I think a rules language would be easier, along the lines of n3.
eric: I ran into problems with
lots of extra blanks, using a rules language.
... When you dive into nested things, it doesn't have an
identity. N3 just runs rules until it ends, and can't be used
for that, because those things are not labeled, so it runs
forever.
david: Maybe something needs to
be added to n3 rules, to retain that context.
... My criticism was not about the functionality of shex, but
the syntax, which is messy. If we could give it a prettier
syntax it might be good for it.
eric: Usually people use app-specific scaffolding that uses specific bnode names, etc.
<scribe> ACTION: Harold to add 1. type arcs for code elements 2. OWL representation of the FHIR code systems (publish them) 3. get valueset information into fhir.ttl . And mention the above on zulip
<trackbot> Sorry, but no Tracker is associated with this channel.
eric: If we want to conneg OWL and Turtle ... when they created the OWL spec they did not distinuish OWL from RDF.
harold: Right now I don't know
exactly what to say in order to get RDF, short of specifically
asking for Turtle.
... Also another task: Both the HAPI and simplifier folks would
be happy to have RDF included, but they don't want to spend the
effort unless we can prove usage. Or do it for them.
harold: Only aware of me using it.
david: Joao sounds interested.
harold: Wikidata person was
excited about it, but got as far as snomed and then said "oh my
god", and stopped.
... They're RDF-ifying the world, bit by bit.
eric: One motivation for them was the gene wiki.
<scribe> ACTION: David to ask Michael van der zel who else is using FHIR/RDF
<trackbot> Sorry, but no Tracker is associated with this channel.
ADJOURNED
This is scribe.perl Revision: 1.152 of Date: 2017/02/06 11:04:15 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00) Present: EricP David_Booth Harold_Solbrig No ScribeNick specified. Guessing ScribeNick: dbooth Inferring Scribes: dbooth WARNING: No date found! Assuming today. (Hint: Specify the W3C IRC log URL, and the date will be determined from that.) Or specify the date like this: <dbooth> Date: 12 Sep 2002 People with action items: david eric harold WARNING: IRC log location not specified! (You can ignore this warning if you do not want the generated minutes to contain a link to the original IRC log.)[End of scribe.perl diagnostic output]