Layers of interoperability, OWL and openEHR

Technical (archive)
1

Dear members of the list,
I’d appreciate your opinions and guidance about a particular topic. As most of you probably know, the work in the ontology domain has been the flagship of semantic interoperability for many projects now, and there is a large amount of researchers active in the field.
I’ve been involved in use of ontologies for semantic interoperability for the first time in 2002, and since then, ontologies have become a frequently pronounced solution for a large set of problems.
However, I have a feeling that the nature of this work creates just a layer in the multilayer interoperability space. Expressing relationships among different entities and doing this in a formal way (OWL) is nice. OWL also allows you to do processing, reasoning on the defined relationships, but unless I’m missing something, this is all about relationships, and concepts. I mean the capabilities of OWL seem to be valid in the relationships is defines.
What about the actual things, data items, entities that OWL links together? I’ve been a proponent of well defined type systems and object hieararchies in healthcare interoperability solutions, since I’ve spent years in the software development side of the domain, and a huge number of issues arise from the developers interpreting losely defined types, or inventing their own types.
Now pinning down concepts either by using terminologies or ontologies is good. It is good to know that two fields on two different data structures are pointing to the same concept. This however, is the beginning of the process. Pointing at the same thing and processing it in the same way are different things. Just because we agree that we are pointing to body temperature in two different documents does not stop us from processing one of them with a double, and the other one with a float.
There is a great deal of information out there expressed in the form of OWL, or other formalisms, but I can’t see this covering all aspects of interoperability, but (no offense) there is a large crowd out there who think they have solved the problem of semantic interoperability. Though it may be an undervaluation of the work, “mappings” are nice, but they don’t ease the rest of the work, where mapped items are processed in different domains.
Are there resources or works that you know of, that try to link type systems in openEHR or other formalisms like 13606 or HL7 to these semantic expressions? How does a DVQuantity instance and an OWL expression play together?

Best Regards
Seref

2

Dear List People,

Another view, and my two (euro) cents, for what they are worth:-

There are many philosophical difficulties in the concept of semantic
interoperability which technology cannot address. Put simply, semantic
interoperability requires an agreement on meaning, and meaning is not a
'thing'. Semantic interoperability requires uses of a system to think
in the same way - or at least in mutually understandable ways - and
informaticians do not (yet) have the power to change the ways people think.

So semantic interoperability is a kind of philosopher's stone. The
search for the original philosopher's stone, which could turn base metal
into gold, simply showed that alchemists misunderstood chemistry and
sub-atomic physics. Maybe the search for semantic interoperability
simply shows that informaticians misunderstand linguistics and the
nature of knowledge.

OK - you can shoot me down now......

Derek.

3

Derek Meyer wrote:

4

Derek,

Shooting?
No.
I agree with you.
And I disagree.

I think that there are clinical informaticians that know, implicitly or explicitly, about semantics, about language and the philosophical aspects.
At least clinicians and nurses do (and most patients and other people) since they communicate using voice, writings and gestures.

The problem is that technicians do not understand semantic interoperability.
And I must say that many informaticians are actually technicians without any understanding of semantics.

Gerard

Dear List People,

Another view, and my two (euro) cents, for what they are worth:-

There are many philosophical difficulties in the concept of semantic
interoperability which technology cannot address. Put simply, semantic
interoperability requires an agreement on meaning, and meaning is not a
‘thing’. Semantic interoperability requires uses of a system to think
in the same way - or at least in mutually understandable ways - and
informaticians do not (yet) have the power to change the ways people think.

So semantic interoperability is a kind of philosopher’s stone. The
search for the original philosopher’s stone, which could turn base metal
into gold, simply showed that alchemists misunderstood chemistry and
sub-atomic physics. Maybe the search for semantic interoperability
simply shows that informaticians misunderstand linguistics and the
nature of knowledge.

OK - you can shoot me down now…

Derek.

– –
Gerard Freriks, MD
Huigsloterdijk 378
2158 LR Buitenkaag
The Netherlands

T: +31 252544896
M: +31 620347088
E: gfrer@luna.nl

Those who would give up essential Liberty, to purchase a little temporary
Safety, deserve neither Liberty nor Safety. Benjamin Franklin 11 Nov 1755

5

Hi Gerard

Who does understand semantic interoperability?

The beauty of human interaction is that we can

get along even without understanding each other.

And we’ll never get computers to understand each

other. So we shouldn’t aim for semantic interoperability,

we should aim for unsemantic interoperability

:wink:

(kudos to the Health IT Nerd)

Grahame

6

Graham,

Exactly.
Somewhere there is a paradox.

Gerard

– –
Gerard Freriks, MD
Huigsloterdijk 378
2158 LR Buitenkaag
The Netherlands

T: +31 252544896
M: +31 620347088
E: gfrer@luna.nl

Those who would give up essential Liberty, to purchase a little temporary
Safety, deserve neither Liberty nor Safety. Benjamin Franklin 11 Nov 1755

7

Who would ever have thought that a technician would have such poetry in their soul - perhaps there is hope for semantic interoperability after all… :slight_smile:

8

Hugh Leslie wrote:

Who would ever have thought that a technician would have such poetry in their
soul - perhaps there is hope for semantic interoperability after all... :slight_smile:
>
>
>
> Who does understand semantic interoperability?
>
> The beauty of human interaction is that we can
>
> get along even without understanding each other.
>
> And we’ll never get computers to understand each
>
> other. So we shouldn’t aim for semantic interoperability,
>
> we should aim for unsemantic interoperability

This is where openEHR gets even more interesting: where Shannon-Weaver
meets Jakobson's six communication functions:

http://en.wikipedia.org/wiki/Roman_Jakobson/The_communication_functions

poetry is the hardest thing to translate but often worth the while.

Discuss!

9

Hugh Leslie wrote:

Who would ever have thought that a technician would have such poetry in their
soul - perhaps there is hope for semantic interoperability after all... :slight_smile:
>
>
>
> Who does understand semantic interoperability?
>
> The beauty of human interaction is that we can
>
> get along even without understanding each other.
>
> And we’ll never get computers to understand each
>
> other. So we shouldn’t aim for semantic interoperability,
>
> we should aim for unsemantic interoperability

Oops that WikiP url was wrong -> the one below works

This is where openEHR gets even more interesting: where Shannon-Weaver
meets Jakobson's six communication functions:

http://en.wikipedia.org/wiki/Roman_Jakobson#The_communication_functions

10

Dear Seref,

Ask yourself the question:
How do we, humans, deal with interoperability?

Do we humans use formally expressed ontologies using OWL.
Do we use rigid formal syntaxes where we use strictly defined formal
terms.
Do wet have to express a measurement in DV-Quantity as Double or
Floating Point with Precision x.
All this is the world of zero's and one's, bits and bytes and IT
industry.

We humans have a vague knowledge of many concepts in our worlds.
We have a very flexible syntax and many, many terms. We even invent
new ones.
It is a chaotic system based on a limited set of rules with emergent
behavior.
We express what we want to document using documents, chapters,
sections, paragraphs, words and characters.
This is the world of documentation, concepts, humans.
This the magnificent world of language, prose and poetry.
Where on the basis of a limited set of rules we can document everything.

It is clear that both worlds (IT and Humans) overlap in certain areas.
But mostly the do not overlap.
Do not mix them up and when you do, we get confused and create monsters.
Both worlds have to stay absolutely orthogonal to each other.

Any interoperability solution where notions, ways of thinking and
expressing, from the IT world with bits and bytes are enforced on
humans, will create problems.
Solutions should start at this human documentation/language level.

The EHR is about documentation of events/facts/thoughts/ideas for
human consumption primarily.
IT-systems should support this. That is all we need for now.
We can try to model real life using the formal, rigid, technical ways.
And create something that doesn't fit the needs of humans or relates
to this human world.
Or we use IT and models to support humans to document what they feel
they need to document.
Humans are not very precise but language works rather efficiently and
well enough.

Modeling knowledge in ontologies is an interesting academic exercise.
Modeling the complex real life is an interesting academic exercise.
But...
Let humans use words freely, either as free text of better from a
common controlled flexible resource (dictionary=coding system/
terminology/classification).
Let humans use words in a syntax (Reference Model) to create freely
all sentences/screens (Templates) they need using agreed documentation
patterns (Archetypes), using tools based on an Archetype Model.

And that for the moment is good enough at this point in time looking
for the Holy Grail called Semantic Interoperability.

Gerard

11

There once was a techie Geek
Who snapped off a email real quieek
Such poetry they said
Without looking in his head
If they had, they'd know it was all cheek

Grahame

12

Can I suggest moving this to the Clinical list? I think it is an
important subject ,and rather dear to my own interests but, as Thomas
pointed out, we are in danger of submerging Seref's original more
technical question.

Any objections?

Ian

Dr Ian McNicoll
office / fax +44(0)141 560 4657
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian@mcmi.co.uk

Clinical Analyst Ocean Informatics ian.mcnicoll@oceaninformatics.com
BCS Primary Health Care Specialist Group www.phcsg.org

13

No, go ahead.

14

An interesting discussion. However, I am aware that humans oftern make
errors, the majority of which is misunderstanding of the words spoken by
another. The commpent "in violent agreement" comes to mind. Even wars have
occured as a result of misunderstanding. Communications and creativity may
be competing terms.

I like this thread of conversation, even tho I don't understand some of the
points.

Ed Hammond

             Gerard Freriks
             <gfrer@luna.nl>
             Sent by: To
             openehr-technical For openEHR technical discussions
             -bounces@openehr. <openehr-technical@openehr.org>
             org cc
                                                                           
                                                                   Subject
             04/22/2009 02:20 Re: Layers of interoperability, OWL
             AM and openEHR
                                                                           
             Please respond to
                For openEHR
                 technical
                discussions
             <openehr-technica
              l@openehr.org>
                                                                           
Dear Seref,

Ask yourself the question:
How do we, humans, deal with interoperability?

Do we humans use formally expressed ontologies using OWL.
Do we use rigid formal syntaxes where we use strictly defined formal
terms.
Do wet have to express a measurement in DV-Quantity as Double or
Floating Point with Precision x.
All this is the world of zero's and one's, bits and bytes and IT
industry.

We humans have a vague knowledge of many concepts in our worlds.
We have a very flexible syntax and many, many terms. We even invent
new ones.
It is a chaotic system based on a limited set of rules with emergent
behavior.
We express what we want to document using documents, chapters,
sections, paragraphs, words and characters.
This is the world of documentation, concepts, humans.
This the magnificent world of language, prose and poetry.
Where on the basis of a limited set of rules we can document everything.

It is clear that both worlds (IT and Humans) overlap in certain areas.
But mostly the do not overlap.
Do not mix them up and when you do, we get confused and create monsters.
Both worlds have to stay absolutely orthogonal to each other.

Any interoperability solution where notions, ways of thinking and
expressing, from the IT world with bits and bytes are enforced on
humans, will create problems.
Solutions should start at this human documentation/language level.

The EHR is about documentation of events/facts/thoughts/ideas for
human consumption primarily.
IT-systems should support this. That is all we need for now.
We can try to model real life using the formal, rigid, technical ways.
And create something that doesn't fit the needs of humans or relates
to this human world.
Or we use IT and models to support humans to document what they feel
they need to document.
Humans are not very precise but language works rather efficiently and
well enough.

Modeling knowledge in ontologies is an interesting academic exercise.
Modeling the complex real life is an interesting academic exercise.
But...
Let humans use words freely, either as free text of better from a
common controlled flexible resource (dictionary=coding system/
terminology/classification).
Let humans use words in a syntax (Reference Model) to create freely
all sentences/screens (Templates) they need using agreed documentation
patterns (Archetypes), using tools based on an Archetype Model.

And that for the moment is good enough at this point in time looking
for the Holy Grail called Semantic Interoperability.

Gerard

15

With Derek's permission, I have started a thread in openEHR-clinical
to continue the clinical /philosophical aspects arising from his reply
to Seref's original post..

"So semantic interoperability is a kind of philosopher's stone...."

Hopefully this will allow Seref to get some sensible technical
solutions to his query and let anyone who fancies a bit of a
philosophical battle to indulge over in openEHR-clinical.

Regards,

Ian

Dr Ian McNicoll
office / fax +44(0)141 560 4657
mobile +44 (0)775 209 7859
skype ianmcnicoll
ian@mcmi.co.uk

Clinical Analyst Ocean Informatics ian.mcnicoll@oceaninformatics.com
BCS Primary Health Care Specialist Group www.phcsg.org

16

I am happy to see responses in the non-technical level too. Well, in case someone has a technical comment regarding binding ontologies to archetypes and openEHR RM objects, I’ll be around :slight_smile:

Kind regards
Seref

17

Seref

I am just starting to look at something in this area.

I am working on how the NHS Logical Record Architecture (LRA) asserts conformance/compliance to external standards. One thing that is required is a semantic mapping between the LRA specifications and the external standard. Initially I am mainly interested in mapping the static models. (Reference models, datatypes, templates, archetypes, etc)

So far I have been interested by the lack of existing work that I have been able to find that asserts the relationship between two specifications – either between two standards, or between a standard and the specification for an implementation. Any suggestions as to where such material may be found would be welcome.

Why has it been hard to find them? Clearly the cost and effort of creating and maintaining such mappings is significant – I suspect that many implementations are developed by rapid prototyping, and the value of maintaining such specifications has not been apparent (maybe because there really was not much value). Maybe part of the reason is that folk have not been actively maintaining specification stacks as systems undergo rapid prototyping. Maybe formal mappings are less important/useful than iterative testing working through examples and scenarios – certainly none of the systems testing that I have seen done in practice has made use of formal conformance statements.

I am looking at OWL and/or ISO 11179 metadata repository specifications as possible formalisms to use – but again would welcome suggestions / comments.

There is a common perception that we have too many overlapping standards in the healthcare space, and that there is a need for consolidation / collaboration and I believe that establishing clear semantic links between specifications may be a way to contribute to convergence. I also believe that it may be a useful contribution to more graceful version management of reference model based specifications within a single standards framework such as HL7v3, 13606, or openEHR

All the best

Charlie

Charlie McCay, charlie@RamseySystems.co.uk
Ramsey Systems Ltd, 23D Dogpole, Shrewsbury, Shropshire SY1 1ES
tel +44 1743 232278 / +44 7808 570172 skype: charliemccay linkedin:charliemccay

18

Dear Seref
As a more technical continuation:
When ontologies and syntaxes are orthogonal the two meet in one place
At that spot on the syntax will refer to a code from a coding system (terminology, classification, code list)
Technically it boils down to how semantically correct and safe can we define this reference?

Ontologies can play a role in the prlduction of codes

Gerard

19

You might find Rahil Qamar's work helpful - http://www.cs.man.ac.uk/~qamarr/

20

Hi Charlie, a couple of good points! Comments are inline.

I am working on how the NHS Logical Record Architecture (LRA) asserts conformance/compliance to external standards. One thing that is required is a semantic mapping between the LRA specifications and the external standard. Initially I am mainly interested in mapping the static models. (Reference models, datatypes, templates, archetypes, etc)

Great starting point. My question is: let’s assume you’ll have the complete mappings tomorrow morning, given to you by someone. For now, let’s say they are expressed in OWL. All the possible mappings for static models you’ve liste are complete. Now, what would you do with them? I’d love to hear your use cases for the situation where you have these mappings.

So far I have been interested by the lack of existing work that I have been able to find that asserts the relationship between two specifications – either between two standards, or between a standard and the specification for an implementation. Any suggestions as to where such material may be found would be welcome.

My first ever involvement with the semantic web in healthcare was in Artemis project, back in 2004, and you may consider taking a look at it. http://www.srdc.metu.edu.tr/webpage/projects/artemis/

Why has it been hard to find them? Clearly the cost and effort of creating and maintaining such mappings is significant – I suspect that many implementations are developed by rapid prototyping, and the value of maintaining such specifications has not been apparent (maybe because there really was not much value). Maybe part of the reason is that folk have not been actively maintaining specification stacks as systems undergo rapid prototyping. Maybe formal mappings are less important/useful than iterative testing working through examples and scenarios – certainly none of the systems testing that I have seen done in practice has made use of formal conformance statements.

I am looking at OWL and/or ISO 11179 metadata repository specifications as possible formalisms to use – but again would welcome suggestions / comments.

Good points. My humble opinion is that, the specifications which are targets of mapping are moving targets, and researchers in the semantic web field seem to be in “art for the sake of art” mode (absolutely no offense) That is, they seem to building capabilities for processing relationships, and processing the actual items involved in relationships is the missing part of the puzzle, or at least that’s my holy grail. I ran into Dr. Dipak Kalra today on my way to grab coffee, and he gave me a couple of nice real life examples, about the things he’d expect from a well established system that has semantic interoperability features. Basically as a clinician he is demanding the capability to ask questions in a single form to heteregenous systems (unless I misunderstood), and this is a good example of a use case, where you need access to real life data, therefore crossing the boundaries of semantic web and ehr related research in both ways. When you create a query in sparql for example, asking for results of a particular treatment for diabetes, you need to use both the established mappings, and the actual values of mapped concepts in different locations and formalisms. I want to hear about suggestions for doing this. Another bright guy we have at CHIME, Matthew Darlison has described similar required his genomics related work, for which he has also not been able to find a satisfying solution.

There is a common perception that we have too many overlapping standards in the healthcare space, and that there is a need for consolidation / collaboration and I believe that establishing clear semantic links between specifications may be a way to contribute to convergence. I also believe that it may be a useful contribution to more graceful version management of reference model based specifications within a single standards framework such as HL7v3, 13606, or openEHR

Having spend about 7 years in healthcare IT now, I have my own reasons to believe in continuing existence of this situation. I have some core use cases, which may provide good starting points for incremental solution attempts to some of the problems we have. For example, as Tony Shannon has written a couple of times, there is ongoing work at CHIME (which should be revealed quite soon) for implementation of fundemental aspects of an openEHR based system, and I am very willing to consider HL7 V3 or 2.x as a message bus for connecting openEHR repositories to other systems. This will require a mapping mechanism, and I would be willing to employ OWL based opportunites here. This would be a controllable use case for tackling the issues which have been described by you also. In case you have other solid use cases you can share, I’d really like to hear about them.

All the best

Many thanks for sharing your opinions,
Kind regards

Seref