RE: Summary of HL7 RDF / W3C COI call: FHIR Ontology Requirements

Rob,
Thanks for the great reference. Do you know whether Pellet and the ICV extension is available for Protégé 5?

Tony


From: Robert Hausam [mailto:rrhausam@gmail.com]
Sent: Saturday, February 07, 2015 12:06 PM
To: Lloyd McKenzie
Cc: Anthony Mallia; Sajjad Hussain; David Booth; w3c semweb HCLS; its@lists.hl7.org
Subject: Re: Summary of HL7 RDF / W3C COI call: FHIR Ontology Requirements

Lloyd, that's certainly correct with the "upper bound", given the conditions that you describe.  If an instance has 5 of "something" when it's declared that it should have 4, then the reasoner can clearly determine that the instance is invalid.  However, using OWA, you can't do this for the "lower bound" of cardinality, as there always may be another "something" out there that the reasoner is not aware of.  I'm sure that we all know all of this, but it definitely makes validating integrity constraints using pure OWL in many cases either difficult or impossible.
I've found this discussion of the issue from Clark&Parsia to be useful:
http://clarkparsia.com/pellet/icv/

This is obviously referring to a proprietary solution (their Pellet reasoner and the ICV extension), and certainly there are other techniques and options available.  But I think this does frame the issue and some potential solutions for it pretty well.
So, getting back to the ontology requirements, I think we clearly will need to be able to use both the open and closed world assumptions, so maybe we should say that we MUST be able to do both? - something like:
MUST: OWL ontology will allow expressions enforcing either closed world or open-world reasoning against instances.

Rob

On Sat, Feb 7, 2015 at 9:07 AM, Lloyd McKenzie <lloyd@lmckenzie.com<mailto:lloyd@lmckenzie.com>> wrote:
Hi Tony,

If you declare an instance has 4 of something, that those instances are disjoint and that the instance is a subclass of those instances that allow only 3 of something, the reasoner *should* declare the instance invalid.  Certainly I was able to get that happening w/ Protege when I used that approach with the RIM.


Lloyd


Lloyd McKenzie
Consultant, Information Technology Services
Gevity Consulting Inc.

 E: lmckenzie@gevityinc.com<mailto:lmckenzie@gevityinc.com>
M: +1 587-334-1110<tel:1-587-334-1110>
W: gevityinc.com<http://gevityinc.com/>

GEVITY
Informatics for a healthier world

CONFIDENTIALITY – This communication is confidential and for the exclusive use of its intended recipients. If you have received this communication by error, please notify the sender and delete the message without copying or disclosing it.

NOTE: Unless explicitly stated otherwise, the opinions and positions expressed in this e-mail do not necessarily reflect those of my employer, my clients nor the organizations with whom I hold governance positions

On Sat, Feb 7, 2015 at 9:03 AM, Anthony Mallia <amallia@edmondsci.com<mailto:amallia@edmondsci.com>> wrote:
Lloyd,
This is the pattern that is used by TopQuadrant in its XSD to OWL conversion and the FHIR generation was shared by Cecil. The advantage of this mechanism is that all subclasses of Patient also are subclasses of the Anonymous Ancestor which is the Class Expression “hasPhoneNumber max 3 PhoneNumber”.

Having done that however the reasoned does not invalidate if there are 4 phone numbers (i.e. Open World).

Tony

From: Lloyd McKenzie [mailto:lloyd@lmckenzie.com<mailto:lloyd@lmckenzie.com>]
Sent: Saturday, February 07, 2015 10:48 AM
To: Sajjad Hussain
Cc: David Booth; w3c semweb HCLS; its@lists.hl7.org<mailto:its@lists.hl7.org>
Subject: Re: Summary of HL7 RDF / W3C COI call: FHIR Ontology Requirements

You can also close the world declaritively.  If I have a Patient with 3 phone numbers, the instance can declare it's a subclass of Patients with an upper bound of 3 on the number of phone numbers. You can do similar things for the vocabulary.  It's verbose, but it works.


Lloyd McKenzie
Consultant, Information Technology Services
Gevity Consulting Inc.

 E: lmckenzie@gevityinc.com<mailto:lmckenzie@gevityinc.com>
M: +1 587-334-1110<tel:1-587-334-1110>
W: gevityinc.com<http://gevityinc.com/>

GEVITY
Informatics for a healthier world

CONFIDENTIALITY – This communication is confidential and for the exclusive use of its intended recipients. If you have received this communication by error, please notify the sender and delete the message without copying or disclosing it.

NOTE: Unless explicitly stated otherwise, the opinions and positions expressed in this e-mail do not necessarily reflect those of my employer, my clients nor the organizations with whom I hold governance positions

On Fri, Feb 6, 2015 at 10:00 PM, Sajjad Hussain <hussain@cs.dal.ca<mailto:hussain@cs.dal.ca>> wrote:
I agree with Lloyd. However, we need to keep in mind that semantic web standard languages especially OWL rely on Open World Assumption (OWA):

http://www.w3.org/TR/2004/REC-owl-guide-20040210/#StructureOfOntologies


For validation purposes, while respecting OWA, it is still possible validate data based on " Scoped Negation as Failure":

https://ai.wu.ac.at/~polleres/publications/poll-etal-2006b.pdf


Best,
Sajjad

******************************************


On 2/6/15 11:29 PM, Lloyd McKenzie wrote:
I expect we'll need to be able to handle both open-world and closed-world versions of the ontology.  Closed-world is essential to validation.  If a profile says something is 1..1 and the instance doesn't have it, then that needs to be flagged as an error, which open-world wouldn't do.  On the other hand, reasoners may well need to operate with some degree of open-world.  The fact something isn't present in the EHR doesn't necessarily mean it isn't true.  I'd be happy for us to include something like this:

SHOULD: OWL ontology should allow expressions enforcing both closed world and open-world reasoning against instances.


Lloyd McKenzie
Consultant, Information Technology Services
Gevity Consulting Inc.

 E: lmckenzie@gevityinc.com<mailto:lmckenzie@gevityinc.com>
M: +1 587-334-1110<tel:1-587-334-1110>
W: gevityinc.com<http://gevityinc.com/>

GEVITY
Informatics for a healthier world

CONFIDENTIALITY – This communication is confidential and for the exclusive use of its intended recipients. If you have received this communication by error, please notify the sender and delete the message without copying or disclosing it.

NOTE: Unless explicitly stated otherwise, the opinions and positions expressed in this e-mail do not necessarily reflect those of my employer, my clients nor the organizations with whom I hold governance positions

On Fri, Feb 6, 2015 at 9:20 PM, David Booth <david@dbooth.org<mailto:david@dbooth.org>> wrote:
Hi Sajjad,

On 02/04/2015 07:12 AM, Sajjad Hussain wrote:
Hi All,

Responding to Action # 2 carried during last call:

http://www.w3.org/2015/02/03-hcls-minutes.html#action02

<http://www.w3.org/2015/02/03-hcls-minutes.html#action02>

I would suggest the following wording for FHIR Ontology Requirement # 11
(http://wiki.hl7.org/index.php?title=FHIR_Ontology_Requirements#11._Enable_Inference

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

11. Enable Inference
(MUST) The FHIR ontology must enable OWL/RDFS inference with
monotonicity and open world assumption [1]
[1] http://www.cs.man.ac.uk/~drummond/presentations/OWA.pdf<http://www.cs.man.ac.uk/%7Edrummond/presentations/OWA.pdf>
<http://www.cs.man.ac.uk/%7Edrummond/presentations/OWA.pdf>

I would expect the closed world assumption to be used quite a lot to  in data validation and perhaps other ways, so I would be uncomfortable having that as a MUST requirement.

David Booth
Best regards,
Sajjad

***************************************************

On 2/3/15 10:45 PM, David Booth wrote:
On today's call we almost finished working out our FHIR ontology
requirements.  Only two points remain to be resolved:
http://wiki.hl7.org/index.php?title=FHIR_Ontology_Requirements


  - Sajjad suggested that the wording of requirement #11 be changed to
be clearer, and agreed to suggest new wording.  Current wording:
"Enable Inference: The FHIR ontology must enable OWL/RDFS inference."

 - Paul Knapp noted that requirement #16 is related to requirement #2,
and suggested that they might be merged.

We did not get to other agenda today.

The full meeting log is here:
http://www.w3.org/2015/02/03-hcls-minutes.html


Thanks!
David Booth


***********************************************************************************
Manage subscriptions - http://www.HL7.org/listservice

View archives - http://lists.HL7.org/read/?forum=its

Unsubscribe - http://www.HL7.org/tools/unsubscribe.cfm?email=lloyd@lmckenzie.com&list=its

Terms of use - http://www.HL7.org/myhl7/managelistservs.cfm?ref=nav#listrules






***********************************************************************************
Manage your subscriptions<http://www.HL7.org/listservice> | View the archives<http://lists.HL7.org/read/?forum=its> | Unsubscribe<http://www.HL7.org/tools/unsubscribe.cfm?email=rrhausam@gmail.com&list=its> | Terms of use<http://www.HL7.org/myhl7/managelistservs.cfm?ref=nav#listrules>



--
Robert Hausam, MD
Hausam Consulting LLC
+1 (801) 949-1556
rrhausam@gmail.com<mailto:rrhausam@gmail.com>