Semantic Web Health Care and Life Sciences Interest Group Teleconference

02 May 2017

See also: IRC log


Ken_Lord, Harold_Solbrig, Rob_Hausam, David_Booth, Sean_Muir, EricP
David Booth


Madrid schedule: Tue May 9, Q4 (4:30-6:00pm local Madrid time)

ken: Can show MDMI tools generating ShExmap

Issue 50: Harmonizing pre-coordinated and compositional terms in the information model

harold: We know how to do post-coordinated. SNOMED International extended the spec to add a specific URI for compositional grammar.

eric: Can it be calculated?

harold: Yes. Fairly simple. But we need to start working out: Existing compositional spec does not have a MIME type. Doesn't tell you what grammar is used.
... Do we need to extend the spec?


harold: When we're converting FHIR to RDF, we could convert a compositional expression to a SNOMED URI. Or we could do what I did in the demo: turn it into an OWL fragment, which is more useful.
... Michael Lawley is the lead on the SNOMED URI work for compositional grammar.
... But there's nothing in the FHIR world that specifies a URI.
... We know that it is SNOMED CT, then could be a post-coord expr, but doesn't say what language it is. FHIR needs the equiv of a MIME type, to indicate it.
... You have to guess from the URI syntax and lack of a number in it.
... If we have the SNOMED use case, we'll soon need the generic use case.

dbooth: Should a request be made to the main FHIR group to add a MIME type indicator?

harold: Yes, at some point.

dbooth: How should we move forward?

harold: First focus on the SNOMED case, and come up with examples of compositional expressions.

<scribe> ACTION: Harold to come up with examples of compositional expression in FHIR [recorded in http://www.w3.org/2017/05/02-hcls-minutes.html#action01]

<trackbot> Created ACTION-83 - Come up with examples of compositional expression in fhir [on Harold Solbrig - due 2017-05-09].

ken: Is CIMI working on this?

harold: IDK
... Previous CIMI datatypes purposely separated a single code to avoid having to parse codes.

<scribe> ... Postponed work on expressions.

rob: The idea was that for any code system there may be a compositional grammar for it.

ken: Is there a goal to have multiple compositional grammars?

harold: The premise that compositional only involves a single code system is pushing your luck. SNOMED takes the philosophy of absorbing everything. But if you are in OWL space nothing prevents you from combining.
... There are grammars that are independent of code systems.
... If we know the code system we could say that that determines the grammar.

eric: When we talk about how there could be multiple compositional grammars, are there multiple already?

harold: OWL and SNOMED. Also HL7 v3 has one of sorts, but don't know whether it creeps into OWL.

rob: FHIR says the compositional grammar must be the one defined by the code system.

harold: Other codes don't fit in well. Also SNOMED has an arcane mechanism for inner concepts, that only works in SNOMED. (Role groups -- Like bnodes.)

eric: SNOMED has already said "Here's what the snomed looks like in OWL, and it has role groups".

harold: Any but the most trivial will be hard to get right, because people don't think that way.

eric: Can we prototype in the FHIR space what it should look like?

dbooth: A replacement for role groups?

eric: Yes.

dbooth: and then propose it to ITHSDO?

harold: Maybe.

eric: Would be interesting, becuase we could use heterogeneous codes, such as RXNORM.
... If we prototyped what SNOMED would look like without role groups, then the thing that would make it interoperate with SNOMED is to use the same relationship name. Then a drug could be identified by RXNORM or others.
... And the use of a more general form of OWL would allow mixing code systems.

dbooth: Is this important enough to work on now?

eric: heterogenous use cases may go away if the compositional grammar allows multiple code systems.

dbooth: I wonder what logic implications might come up, with multiple code systems.

eric: Different granularity.
... Use cases should show what rules would be fired differently if you have one or the other.

dbooth: Use cases would help.

Protege does not show bnodes


eric: They understand the problem.
... There was some conversation about how to find funding for them. Either programmer or money.

dbooth: Eric, are you still planning to pursue the turtle parser problem?

(lost eric from the call?)

FHIR RDF ontology URLs do not resolve


<scribe> ACTION: Harold to initiate discussion with Grahame about adding conneg and a variable for ont version. [recorded in http://www.w3.org/2017/05/02-hcls-minutes.html#action02]

<trackbot> Created ACTION-84 - Initiate discussion with grahame about adding conneg and a variable for ont version. [on Harold Solbrig - due 2017-05-09].

harold: not done yet

How to handle required/extensible/example in shex #47

Harold: This is about code system.
... Required is ok, because we can put the values into shex. But what to do in shex if it is extensible?
... THis problem is not specific to shex. SHould look at other validators, what they do.

rob: It will allow other codes, but they should be within the same code system.
... Ideally if you give a code that's part of the extension space and its a code that conveys something that is already in the code system, then you should use it. Ideally validator should warn if you use one that is already there (or in a broader form).

dbooth: So if the code system is extensible, then the validator should allow anything?

rob: If the specified valueset has a code for blue, and you want to send a code for navy, you are not allowed to, because the meaning is in the specified valueset. But not everyone agrees with that.

eric: THat helps interop.

dbooth: So under that interpretation, then extensibility is NOT for finer granularity.

rob: Correct. It is for additional meaning not already covered.

dbooth: Should we close this issue regarding shex validation, because we cannot do any validation for 'extensible'?

harold: yes. But we should validate 'required'.

dbooth: Should anything be done regarding 'example'?

rob: all examples should be valid.

harold: We cannot check them anyway, because there is nothing to check them against.

AGREED: Close issue 47 - no shex validation needed for extensible/example.


Summary of Action Items

[NEW] ACTION: Harold to come up with examples of compositional expression in FHIR [recorded in http://www.w3.org/2017/05/02-hcls-minutes.html#action01]
[NEW] ACTION: Harold to initiate discussion with Grahame about adding conneg and a variable for ont version. [recorded in http://www.w3.org/2017/05/02-hcls-minutes.html#action02]

Summary of Resolutions

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.152 (CVS log)
$Date: 2017/05/02 16:07:14 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
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: s/Useful to CIMI also/Is CIMI working on this/
Succeeded: s/David Booth/David_Booth/
Succeeded: s/Sean Muir/Sean_Muir/
Present: Ken_Lord Harold_Solbrig Rob_Hausam David_Booth Sean_Muir EricP
No ScribeNick specified.  Guessing ScribeNick: dbooth
Inferring Scribes: dbooth
Found Date: 02 May 2017
Guessing minutes URL: http://www.w3.org/2017/05/02-hcls-minutes.html
People with action items: harold

[End of scribe.perl diagnostic output]