See also: IRC log
Topic Example of connecting FHIR/RDF to external ontology
Gopi: Played with FHIR RDF allergy intolerance example. Created simple prototype. Not using using SNOMED, BFO etc, but it is using dbpedia ont.
(gopi shows screen demo)
gopi: Used SNOMED superclass/subclass info for Hives.
(also showed music decision support using dbpedia data)
gopi: next step is to dev a more useful clinical example
dbooth: planning to connect this to dentistry ont or data?
gopi: Yes, but still investigating.
eric: I can see a couple of
strategies for using subsumption, either for research or
decision support.
... Tempted to monitor temperature of someone during
immunosuppressants, to see if they're having a reaction.
... Tooling is working when you can match instance data with
terms that exactly match. But the tooling is more valuable when
it matches something higher in the subsumption hierarchy.
gopi: SNOMED does not cover dentistry well. Need to use SNODENT.
joe: re there other external ontologies to connect to based on FHIR/RDF?
dbooth: I don't know of any that are based on FHIR RDF.
gopi: Might try to connect to OBO.
(skip this topic; rob not available)
sarah: Long-time snomed user
harold: Good talk with Michael Lawley.
<scribe> ... Completed a week-long mtg about extending snomed DL and how to publish it in OWL.
UNKNOWN_SPEAKER: There's a weird
thing in SNOMED called Role Groups.
... If your compositional grammar is "close to user", then
somebody needs to turn it into a corresponding OWL
expression.
... There's talk of standing up an open server to do
that.
... For the moment we should wait until things are more settled
in snomed.
... Talking about post-coordinating expressions.
... The only example of snomed post-coord on the FHIR STU3 web
site uses one of the few examples that does not involve role
groups, i.e., laterality.
... That was trivial to convert into OWL.
... But when we involve body sites and etiologies, the
compositional grammer must be "close to user", so one must be
clever about getting it into OWL.
... There's discussion about snomed hosting a server (open
source) that will do this sort of conversion for us.
... Not many are using compositional grammar. Best to move
through this methodically.
Discussion on zulip: https://chat.fhir.org/#narrow/stream/ontology/subject/SNOMED.20Expressions
Turtle example: http://build.fhir.org/observation-example-bmd.ttl.html
Same example in JSON: http://build.fhir.org/observation-example-bmd.json.html
harold: Suggest revisiting this
in a month or two.
... Michael Lawley and Linda Bird may help us.
rob: Harold and I planned to work on it, but haven't yet.
harold: I'm working on another
project that uses the FHIR ontology, and incorporates into
I2B2, and that needs required value lists.
... Would be nice to have those done.
... Grahahe proposed that we go ahead and do the conversions in
the build system so that folks can see them, and then the Vocab
group can discuss whether to include them in the OWL that is
released.
... Will do in the next few weeks.
harold: FHIR RDF gets us a lot
closer to the I2B2 goal then we would have. I2B2 is what RDF
might have been if it had not been thought about for long
time.
... There's a subject, and a predicate called 'code'. Now they
have a modifier and a value field.
... If we can make this work we might make I2B2 more
portable.
... Need to add to agenda: In the TTL we have deeply nested
bnodes in some spots. If you're going to ref the units, you
need a whole chain of predicates
... But I2B2 doesn't allow that. ARe there ways that we can
make predicates that represent a long chain of predicates.
ken: We worked on conversion of FHIR and CCDA to I2B2. Issue: difference between a fact-based approach vs observation-based approach.
harold: Working with Sean Murphy.
He pointed out the partial relation between I2B2 and RDF.
... Having the ont and data available in RDF would have been
daunting. FHIR RDF made it only half daunting.
<scribe> ACTION: dbooth to ping Ken about presenting about I2B2 mapping experience [recorded in http://www.w3.org/2017/06/13-hcls-minutes.html#action01]
<trackbot> Error finding 'dbooth'. You can review and register nicknames at <http://www.w3.org/2014/HCLS/track/users>.
ken: Suggest talking to Jeff Kwon.
<inserted> Rob's fork of Harold's demo: https://github.com/rhausam/BLENDINGFHIRandRDF
dbooth: can we highlight and promote your demo more now?
harold: The web site and demo are
backing the paper that we submitted.
... Next steps include pursuing ont other than snomed.
... Got snomed people looking at the compositional grammar and
how we've done this.
... Grahame has expressed an interest in extending it, using
model def work and 11179.
dbooth: Would like to see it prepared as a demo.
harold: It's not far from that, but it's supporting a paper right now, so it should be forked instead of modified directly.
Here is a recording of Harold's demo; https://drive.google.com/file/d/0B_OSV7VKS60hOUpTNjRfb285VDg/view?usp=sharing
harold: If this were used in
production it would need to be connected to a server instead of
protege.
... Michael Lawley thought it would be relatively trivial to
make snorocket handle everything that's in there.
... Also a bunch of snomed folks are using the Elk reasoner,
because it's slightly faster.
<scribe> ACTION: Rob to fork and enhance Harold's demo as a tutorial. [recorded in http://www.w3.org/2017/06/13-hcls-minutes.html#action02]
<trackbot> Created ACTION-88 - Fork and enhance harold's demo as a tutorial. [on Rob Frost - due 2017-06-20].
harold: protege does not show info on bnodes.
Eric: No progress yet. I didn't think it would be easy to solve. Harold: Depends where the problem is. Need to keep the pressure on them though. Eric: The OWL API is mutable. Harold: You might try skolemizing the bnodes and see if it deals with it better if it has names. Eric: We cannot put skolem URIs into the OWL itself, but we can put them into the data.
<scribe> ACTION: Eric to try skolemizing bnodes to see if that helps. That will give more info. Then ping protege developers again. [recorded in http://www.w3.org/2017/06/13-hcls-minutes.html#action03]
<trackbot> Created ACTION-89 - Try skolemizing bnodes to see if that helps. that will give more info. then ping protege developers again. [on Eric Prud'hommeaux - due 2017-06-20].
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) Succeeded: i|can we highlight|Rob's fork of Harold's demo: https://github.com/rhausam/BLENDINGFHIRandRDF Present: Harold_Solbrig David_Booth Gopi Joseph_Quinn Rob_Hausam Sarah_Ryan EricP Ken_Lord No ScribeNick specified. Guessing ScribeNick: dbooth Inferring Scribes: dbooth Found Date: 13 Jun 2017 Guessing minutes URL: http://www.w3.org/2017/06/13-hcls-minutes.html People with action items: dbooth eric rob[End of scribe.perl diagnostic output]