side technical discussion on 11179 and DL ontology

classic Classic list List threaded Threaded
9 messages Options
Reply | Threaded
Open this post in threaded view
|

side technical discussion on 11179 and DL ontology

Paul S Prueitt


Rebecca,

First a review of the discussion between you and Thomas so far... and then
some questions and comments.

You said

" I mentioned that
The Cancer Bioinformatics Grid (caBIG)... has a ISO/IEC
111179 compliant registry that links Data Element Concepts, Valid Values,
Representation Terms, and other
administered components to concepts in a DL (Descriptive Logic) based
ontology - NCI Thesaurus. NCI has made some extensions
to the ISO/IEC standard - which I think they have proposed back to the
standards organization for linking to an ontology in this fashion."


On the link you provided

http://ncicb.nci.nih.gov/NCICB/infrastructure/cacore_overview/cadsr

It is mentioned that

The caCORE objects are represented by UML Models..

Question:  but is UML a Descriptive Logic ontology?  I would not be
surprised if DL can be, is, applied to UML elements.  But, as in the fine
vocabulary of the 11179, the vocabulary of UML is not (in my understanding)
set up perfectly to arrange concepts into hierarchical inheritance trees
with logical "axioms" asserting inferential properties such an transitive
and reflective assignments, as well as the class sub-class definitions.  (I
hope I said that ok.)  In fact this translation issue is a big deal right
now. Right?

You mention an almost complete paper describing "semantic linkage".  And by
"semantic linkage" you mean precisely a mapping from the 11179 to a DL based
logic (say Protege Frames (CLIPS) or OWL).

As you might know, I (and others) would express some concern about the use
of that language, as the real task is structural interoperability between
finite state machines.  Right?

Meaning is properly applied to fixed intended interpretations, of course
this is true,  but a deeper connotation for the term "meaning" comes only
with differential interpretation and in this case the DL ontologies have
both practical and theoretical problems with computability, as has been
pointed out by many.  I hope I am being clear and non-controversal in
pointing this out.


You ask Thomas

"if what you are proposing is to actually implement the ISO standard as
a frame system in protege - I think that would be very useful"


Question 1: what is the DL based ontology that (you mentioned) is now used
to link, (semantically link?) the ISO/IEC 11179 standard to?  Is this not
Protege Frames?  Is it UML equiped with logic?  I am unclear about what you
are talking about.

Question 2:  When one is talking about 11170 compliance, what does this mean
precisely (to you)?

http://isotc.iso.org/livelink/livelink/fetch/2000/2489/Ittf_Home/PubliclyAva
ilableStandards.htm

lists the many ISO standards.

the 11170 has six parts (around 400 pages)..

Fortunately there is a lot of repetition in these parts and within those
pages.

In reading through these pages, I was very impressed with the degree to
which (certain types of) data standardization could be achieved within this
standard.  It is comprehensive and clear.

I am working on a "semantic linkage" between OASIS SOA-IM where again a
quite general understanding of a data modeling framework is available.
Service Oriented Architecture Information Model is far simpler that the
11179, but the key issues are almost the same.

One of the Protege forum contributors, Andrea, made a very proper suggestion
(not that I knew what was proper until I read his note)

http://www.bcngroup.org/beadgames/generativeMethodology/166.htm

I was searching for an answer (in fact I was also seraching to ask the
question correctly), when a stranger (but now a friend) just nailed it!!!!


I should say, not that it makes a lot of difference, that Andrea's use of
the term "non-monotonic" is narrow...  (non-monotonic often means not only
does new information override previous assertions of truth but it does so in
a way in which the location of that which is over written is uncertain.)
But this narrow definition of non-monotonic is wonderfully useful in doing
as much as possible to make structural mappings between things UML, 11179,
and SIA-IM like to things that are like OWL and Protege Frames (ie the
Descriptive Logic type ontologies).

I would appreciate general comments and corrections to my opinions, as we
are all trying to do something that is not fully understood.  So I (as we
all will) will make mistakes ..










-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Andrea Proli

Paul, Rebecca and all,
UML has no attached DL, and it has no precise specification at all of the  
meaning of its elements. It is only intended to defines a graphical  
notation and nothing more. Actually, UML has 9 kinds of diagrams which are  
intended to model both the static aspects and the dynamic behavior of a  
system, but I suppose you guys refer to class diagrams for static modeling  
only here.
Anyway, one can give the semantics he prefers to an UML (class) diagram.  
This is a point of strong criticism. To understand why, consider that UML  
allows to draw objects which are instances of a class, but it does not  
tell, for example, whether an object can be an instance of multiple  
classes and/or can change the class(es) it is an instance of during time  
(in set theory: "can an element be included in more than one set? can an  
element belong to different sets in different time instants?"). This is  
often considered to obviously hold in the Knolwedge Representation  
community where "class" means "set", but it is not so in the context of  
object-oriented programming, where for efficiency reasons it is assumed  
that objects are instances of a single class and that the class they are  
instance of never changes until the object is destroyed (ok, with  
exceptions, say Smalltalk).
So, UML is pretty good for documenting software design, where many  
contextual assumptions are implicitly made, but personally I would not  
rely on UML if I were to define an unambiguous standard specification.
Best regards,

Andrea

On Mon, 13 Feb 2006 05:20:41 +0100, Paul S Prueitt  
<[hidden email]> wrote:

>
> Rebecca,
>
> First a review of the discussion between you and Thomas so far... and  
> then
> some questions and comments.
>
> You said
>
> " I mentioned that
> The Cancer Bioinformatics Grid (caBIG)... has a ISO/IEC
> 111179 compliant registry that links Data Element Concepts, Valid Values,
> Representation Terms, and other
> administered components to concepts in a DL (Descriptive Logic) based
> ontology - NCI Thesaurus. NCI has made some extensions
> to the ISO/IEC standard - which I think they have proposed back to the
> standards organization for linking to an ontology in this fashion."
>
>
> On the link you provided
>
> http://ncicb.nci.nih.gov/NCICB/infrastructure/cacore_overview/cadsr
>
> It is mentioned that
>
> The caCORE objects are represented by UML Models..
>
> Question:  but is UML a Descriptive Logic ontology?  I would not be
> surprised if DL can be, is, applied to UML elements.  But, as in the fine
> vocabulary of the 11179, the vocabulary of UML is not (in my  
> understanding)
> set up perfectly to arrange concepts into hierarchical inheritance trees
> with logical "axioms" asserting inferential properties such an transitive
> and reflective assignments, as well as the class sub-class definitions.  
> (I
> hope I said that ok.)  In fact this translation issue is a big deal right
> now. Right?
>
> You mention an almost complete paper describing "semantic linkage".  And  
> by
> "semantic linkage" you mean precisely a mapping from the 11179 to a DL  
> based
> logic (say Protege Frames (CLIPS) or OWL).
>
> As you might know, I (and others) would express some concern about the  
> use
> of that language, as the real task is structural interoperability between
> finite state machines.  Right?
>
> Meaning is properly applied to fixed intended interpretations, of course
> this is true,  but a deeper connotation for the term "meaning" comes only
> with differential interpretation and in this case the DL ontologies have
> both practical and theoretical problems with computability, as has been
> pointed out by many.  I hope I am being clear and non-controversal in
> pointing this out.
>
>
> You ask Thomas
>
> "if what you are proposing is to actually implement the ISO standard as
> a frame system in protege - I think that would be very useful"
>
>
> Question 1: what is the DL based ontology that (you mentioned) is now  
> used
> to link, (semantically link?) the ISO/IEC 11179 standard to?  Is this not
> Protege Frames?  Is it UML equiped with logic?  I am unclear about what  
> you
> are talking about.
>
> Question 2:  When one is talking about 11170 compliance, what does this  
> mean
> precisely (to you)?
>
> http://isotc.iso.org/livelink/livelink/fetch/2000/2489/Ittf_Home/PubliclyAva
> ilableStandards.htm
>
> lists the many ISO standards.
>
> the 11170 has six parts (around 400 pages)..
>
> Fortunately there is a lot of repetition in these parts and within those
> pages.
>
> In reading through these pages, I was very impressed with the degree to
> which (certain types of) data standardization could be achieved within  
> this
> standard.  It is comprehensive and clear.
>
> I am working on a "semantic linkage" between OASIS SOA-IM where again a
> quite general understanding of a data modeling framework is available.
> Service Oriented Architecture Information Model is far simpler that the
> 11179, but the key issues are almost the same.
>
> One of the Protege forum contributors, Andrea, made a very proper  
> suggestion
> (not that I knew what was proper until I read his note)
>
> http://www.bcngroup.org/beadgames/generativeMethodology/166.htm
>
> I was searching for an answer (in fact I was also seraching to ask the
> question correctly), when a stranger (but now a friend) just nailed  
> it!!!!
>
>
> I should say, not that it makes a lot of difference, that Andrea's use of
> the term "non-monotonic" is narrow...  (non-monotonic often means not  
> only
> does new information override previous assertions of truth but it does  
> so in
> a way in which the location of that which is over written is uncertain.)
> But this narrow definition of non-monotonic is wonderfully useful in  
> doing
> as much as possible to make structural mappings between things UML,  
> 11179,
> and SIA-IM like to things that are like OWL and Protege Frames (ie the
> Descriptive Logic type ontologies).
>
> I would appreciate general comments and corrections to my opinions, as we
> are all trying to do something that is not fully understood.  So I (as we
> all will) will make mistakes ..
>
>
>
>
>
>
>
>
>
>



--
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Dirk Riehle-2


>UML has no attached DL, and it has no precise specification at all of the
>meaning of its elements. It is only intended to defines a graphical
>notation and nothing more. Actually, UML has 9 kinds of diagrams which are

UML certainly has higher aspirations than just being a graphical
notation. For one, it certainly wants to be a fully executable
language one day and people are working on it.

There is no reason to leave a major language specified unambiguously
deliberately. It just takes time, like any large specification effort.

>This is
>often considered to obviously hold in the Knolwedge Representation
>community where "class" means "set", but it is not so in the context of
>object-oriented programming, where for efficiency reasons it is assumed
>that objects are instances of a single class and that the class they are

I think you got it backwards. Despite its heritage in programming
languages, UML was designed to be expressive, hence the large number
of different modeling techniques it offers.

OWL and the like have been designed with computability issues in mind
(rather than giving the modeler what he needs). The whole extrinsic
vs intrinsic definition of classes discussion shows that nicely. If
you want an inferencing engine, you need a set-theory like approach.
Since UML users typically don't need that, they have not constrained by this.

On the specific issue of an object being an instance of multiple
classes: I think direct instance-of is overused, massively. What is
missing is more use of a concept like is-role-of, and that would do
away with these frequently incomprehensibly deep class hierarchies.

Dirk



Dirk Riehle, ph: +49 172 184 8755, web: http://www.riehle.org
Interested in wiki research? Please see http://www.wikisym.org!

-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Andrea Proli

Hi Dirk,
UML was intentionally designed without an unambiguous formal semantics,  
and after 8 years it still does not have one: indeed, UML is defined by  
means of itself, i.e. it is specified by its metamodel, which in turn is  
expressed in UML. This reflect a precise conscious design decision not to  
give it a formal semantics, and I did not say that this is necessarily  
bad, but just that it is a strong point of criticism; anyway, I am really  
glad to hear that someone is eventually working on that. As long as no  
formal, standard, and agreed-upon specification of its semantics is given,  
however, UML remains a mere graphical notation. As to the "expressiveness"  
issue, no quantification of "expressiveness" is possible unless a formal  
framework that denotes *what* exactly a UML model expresses, and how  
relationships, classes, and objects exactly relate to one another, is  
provided.

The reason why you perceive that the instance-of relationship is overused  
and you suggest to replace it with is-role-of properties, in my opinion,  
is not that you don't really need the semantics of the instance-of  
relationship, but that you want instance-of relationships to behave like  
is-role-of properties in ordinary programs, i.e. being able to change  
dynamically and possibly assume several values. This is probably due to  
the fact that you are deeply involved in the Software Engineering  
discipline, and I can understand your point of view because I also was a  
few time ago. But still, replacing instance-of with is-role-of properties  
is just a trick, an implementation issue that arises because you have  
already made a committment to the execution platform (the logical model)  
in which you are going to translate your conceptual models: that of an  
object-oriented programming language. Actually, you don't want to give up  
the semantics of instance-of, you just want to overcome the limitations  
which are imposed by those environments because they designed to maximize  
efficiency.

Please note, that I am not claiming that object-oriented programming  
languages are something "improper", just that they have very specific  
goals, and "expressive" conceptual modeling languages should have a  
denotational, set-based, model-theoretic semantics in order to be regarded  
as such. And this applies to UML, if its expectations include, as you say,  
not to be just a graphical notation.

Anyway, Dirk, in case you felt my post was rude or offending, I  apologize  
for that: don't feel attacked. I was only trying to expose my point of  
view, maybe I did it with too much strength ;)
Friendly,

Andrea


On Mon, 13 Feb 2006 12:46:15 +0100, Dirk Riehle <[hidden email]> wrote:

>
>
>> UML has no attached DL, and it has no precise specification at all of  
>> the
>> meaning of its elements. It is only intended to defines a graphical
>> notation and nothing more. Actually, UML has 9 kinds of diagrams which  
>> are
>
> UML certainly has higher aspirations than just being a graphical
> notation. For one, it certainly wants to be a fully executable
> language one day and people are working on it.
>
> There is no reason to leave a major language specified unambiguously
> deliberately. It just takes time, like any large specification effort.
>
>> This is
>> often considered to obviously hold in the Knolwedge Representation
>> community where "class" means "set", but it is not so in the context of
>> object-oriented programming, where for efficiency reasons it is assumed
>> that objects are instances of a single class and that the class they are
>
> I think you got it backwards. Despite its heritage in programming
> languages, UML was designed to be expressive, hence the large number
> of different modeling techniques it offers.
>
> OWL and the like have been designed with computability issues in mind
> (rather than giving the modeler what he needs). The whole extrinsic
> vs intrinsic definition of classes discussion shows that nicely. If
> you want an inferencing engine, you need a set-theory like approach.
> Since UML users typically don't need that, they have not constrained by  
> this.
>
> On the specific issue of an object being an instance of multiple
> classes: I think direct instance-of is overused, massively. What is
> missing is more use of a concept like is-role-of, and that would do
> away with these frequently incomprehensibly deep class hierarchies.
>
> Dirk
>
>
>
> Dirk Riehle, ph: +49 172 184 8755, web: http://www.riehle.org
> Interested in wiki research? Please see http://www.wikisym.org!
>
> -------------------------------------------------------------------------
> To unsubscribe go to http://protege.stanford.edu/community/subscribe.html
>



--
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Dirk Riehle-2

Hi Andrea,

I'd be surprised to hear anyone at the OMG ever said that UML was
deliberately designed to be imprecise.  The OMG is driven to a large
extent by the tool vendors, and lack of interoperability hinders
their business, so they will eventually fix it. Academic groups, for
one, have long worked on a formalization of the UML core. I guess it
will remain an open issue whether it really needs denotational semantics :-)

I work with is-role-of relationships in addition to instance-of
relationship because it lets me more precisely capture the semantics
of the domain I'm working on. It really doesn't have much to do with
the traditional database notion of type migration or dynamically
attaching objects. Please note that I'm not doing away with
instance-of (how could I), I'm just saying it is overused.

Take a typical example. If I model a Party (or Person) concept in the
banking domain, I could have a large class graph with various shades
of Customer, Debtor, Guarantor, Investor, classes etc. Or I could
have a Person class in the middle and define concepts of Customer,
Debtor, etc as roles of a Person around it. I'd argue that roles
enrich my modeling vocabulary substantially (to the extent that I
believe it should be part of modeling languages) and let me represent
a domain more effectively.

I wasn't offended by our original post. I apologize if I came across
like that---my English is sometimes more curt than I wished it would be.

Dirk



At 13.02.2006, you wrote:

>Hi Dirk,
>UML was intentionally designed without an unambiguous formal semantics,
>and after 8 years it still does not have one: indeed, UML is defined by
>means of itself, i.e. it is specified by its metamodel, which in turn is
>expressed in UML. This reflect a precise conscious design decision not to
>give it a formal semantics, and I did not say that this is necessarily
>bad, but just that it is a strong point of criticism; anyway, I am really
>glad to hear that someone is eventually working on that. As long as no
>formal, standard, and agreed-upon specification of its semantics is given,
>however, UML remains a mere graphical notation. As to the "expressiveness"
>issue, no quantification of "expressiveness" is possible unless a formal
>framework that denotes *what* exactly a UML model expresses, and how
>relationships, classes, and objects exactly relate to one another, is
>provided.
>
>The reason why you perceive that the instance-of relationship is overused
>and you suggest to replace it with is-role-of properties, in my opinion,
>is not that you don't really need the semantics of the instance-of
>relationship, but that you want instance-of relationships to behave like
>is-role-of properties in ordinary programs, i.e. being able to change
>dynamically and possibly assume several values. This is probably due to
>the fact that you are deeply involved in the Software Engineering
>discipline, and I can understand your point of view because I also was a
>few time ago. But still, replacing instance-of with is-role-of properties
>is just a trick, an implementation issue that arises because you have
>already made a committment to the execution platform (the logical model)
>in which you are going to translate your conceptual models: that of an
>object-oriented programming language. Actually, you don't want to give up
>the semantics of instance-of, you just want to overcome the limitations
>which are imposed by those environments because they designed to maximize
>efficiency.
>
>Please note, that I am not claiming that object-oriented programming
>languages are something "improper", just that they have very specific
>goals, and "expressive" conceptual modeling languages should have a
>denotational, set-based, model-theoretic semantics in order to be regarded
>as such. And this applies to UML, if its expectations include, as you say,
>not to be just a graphical notation.
>
>Anyway, Dirk, in case you felt my post was rude or offending, I  apologize
>for that: don't feel attacked. I was only trying to expose my point of
>view, maybe I did it with too much strength ;)
>Friendly,
>
>Andrea
>
>
>On Mon, 13 Feb 2006 12:46:15 +0100, Dirk Riehle <[hidden email]> wrote:
>
> >
> >
> >> UML has no attached DL, and it has no precise specification at all of
> >> the
> >> meaning of its elements. It is only intended to defines a graphical
> >> notation and nothing more. Actually, UML has 9 kinds of diagrams which
> >> are
> >
> > UML certainly has higher aspirations than just being a graphical
> > notation. For one, it certainly wants to be a fully executable
> > language one day and people are working on it.
> >
> > There is no reason to leave a major language specified unambiguously
> > deliberately. It just takes time, like any large specification effort.
> >
> >> This is
> >> often considered to obviously hold in the Knolwedge Representation
> >> community where "class" means "set", but it is not so in the context of
> >> object-oriented programming, where for efficiency reasons it is assumed
> >> that objects are instances of a single class and that the class they are
> >
> > I think you got it backwards. Despite its heritage in programming
> > languages, UML was designed to be expressive, hence the large number
> > of different modeling techniques it offers.
> >
> > OWL and the like have been designed with computability issues in mind
> > (rather than giving the modeler what he needs). The whole extrinsic
> > vs intrinsic definition of classes discussion shows that nicely. If
> > you want an inferencing engine, you need a set-theory like approach.
> > Since UML users typically don't need that, they have not constrained by
> > this.
> >
> > On the specific issue of an object being an instance of multiple
> > classes: I think direct instance-of is overused, massively. What is
> > missing is more use of a concept like is-role-of, and that would do
> > away with these frequently incomprehensibly deep class hierarchies.
> >
> > Dirk
> >
> >
> >
> > Dirk Riehle, ph: +49 172 184 8755, web: http://www.riehle.org
> > Interested in wiki research? Please see http://www.wikisym.org!
> >
> > -------------------------------------------------------------------------
> > To unsubscribe go to http://protege.stanford.edu/community/subscribe.html
> >
>
>
>
>--
>Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
>-------------------------------------------------------------------------
>To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Paul Prueitt


Andrea and Dirk,


Excellent discussion.

I suggest that the

"The capture of semantics" has a meaning at two levels, and in two very
different contexts.

It could mean "some" annotation between UML or a object oriented data model
and some model of the meaning of data/information elements.  In this sense
it is proper to say, as Rebecca was saying, that a semantic linkage might be
created as a bridge between a data model and an information model.

It is also proper to say, as some suggest, that DL-based logic and ontology
specifications can be used to index data that is organized by registries
like those that are compliant with ISO 11179.  Said in a slightly different
way, the BioPAX OWL (Full, I believe) rendering of a information model for
integrating all gene and cell signal pathway data (from very diverse
bioinformatics projects) my be able to index all existing bioinformatics
data sources using a DL-type (class - subclass) ontology.  (As I have
pointed out, this has a huge benefit but some surprises.)

Most know my position regarding the second context.  There are some
consequences to assuming that the folks at OMG (Object Management Group)
will one day be able to make UML "expressive" in the sense defined by
DLbased ontology.

Expressive of "what"?  Two cases exist.

If the data space is stable, well defined with no ambiguity and not complex
than we all feel that translations between ISO 11179 or SAO IM or other data
object standard, and DL-type ontology will be figured out this year!!!.  But
this only means that a translation between two theories related to data
organization will then exist.  John Sowa's often talked about "lattice of
theories" comes to mind.

Andrea's solution to the translation of a subset of the SOA IM

http://www.bcngroup.org/beadgames/generativeMethodology/163.htm

is being widely discussed in circles (outside these forums), and is widely
seen as being "correct".


But what about "web service" discovery, where the discovery effort is being
made in a dynamic environment having lots of novelty and is time
constrained.

Scenario:  I (a federal agency) am trying to discovery a set of services
related to how timely response to a natural disaster can be made.  "I" am a
EU country so these service discovery requests have to be contextualized not
only into different human languages but with cultural contexts that vary
even if there is an absolutely agreed to and non ambiguous interlingua
representation of both relevant DL ontologies and data object information
models.






-----Original Message-----
From: [hidden email]
[mailto:[hidden email]]On Behalf Of Dirk
Riehle
Sent: Monday, February 13, 2006 9:09 AM
To: [hidden email]
Subject: [protege-discussion] Re: side technical discussion on 11179 and
DL ontology



Hi Andrea,

I'd be surprised to hear anyone at the OMG ever said that UML was
deliberately designed to be imprecise.  The OMG is driven to a large
extent by the tool vendors, and lack of interoperability hinders
their business, so they will eventually fix it. Academic groups, for
one, have long worked on a formalization of the UML core. I guess it
will remain an open issue whether it really needs denotational semantics :-)

I work with is-role-of relationships in addition to instance-of
relationship because it lets me more precisely capture the semantics
of the domain I'm working on. It really doesn't have much to do with
the traditional database notion of type migration or dynamically
attaching objects. Please note that I'm not doing away with
instance-of (how could I), I'm just saying it is overused.

Take a typical example. If I model a Party (or Person) concept in the
banking domain, I could have a large class graph with various shades
of Customer, Debtor, Guarantor, Investor, classes etc. Or I could
have a Person class in the middle and define concepts of Customer,
Debtor, etc as roles of a Person around it. I'd argue that roles
enrich my modeling vocabulary substantially (to the extent that I
believe it should be part of modeling languages) and let me represent
a domain more effectively.

I wasn't offended by our original post. I apologize if I came across
like that---my English is sometimes more curt than I wished it would be.

Dirk



At 13.02.2006, you wrote:

>Hi Dirk,
>UML was intentionally designed without an unambiguous formal semantics,
>and after 8 years it still does not have one: indeed, UML is defined by
>means of itself, i.e. it is specified by its metamodel, which in turn is
>expressed in UML. This reflect a precise conscious design decision not to
>give it a formal semantics, and I did not say that this is necessarily
>bad, but just that it is a strong point of criticism; anyway, I am really
>glad to hear that someone is eventually working on that. As long as no
>formal, standard, and agreed-upon specification of its semantics is given,
>however, UML remains a mere graphical notation. As to the "expressiveness"
>issue, no quantification of "expressiveness" is possible unless a formal
>framework that denotes *what* exactly a UML model expresses, and how
>relationships, classes, and objects exactly relate to one another, is
>provided.
>
>The reason why you perceive that the instance-of relationship is overused
>and you suggest to replace it with is-role-of properties, in my opinion,
>is not that you don't really need the semantics of the instance-of
>relationship, but that you want instance-of relationships to behave like
>is-role-of properties in ordinary programs, i.e. being able to change
>dynamically and possibly assume several values. This is probably due to
>the fact that you are deeply involved in the Software Engineering
>discipline, and I can understand your point of view because I also was a
>few time ago. But still, replacing instance-of with is-role-of properties
>is just a trick, an implementation issue that arises because you have
>already made a committment to the execution platform (the logical model)
>in which you are going to translate your conceptual models: that of an
>object-oriented programming language. Actually, you don't want to give up
>the semantics of instance-of, you just want to overcome the limitations
>which are imposed by those environments because they designed to maximize
>efficiency.
>
>Please note, that I am not claiming that object-oriented programming
>languages are something "improper", just that they have very specific
>goals, and "expressive" conceptual modeling languages should have a
>denotational, set-based, model-theoretic semantics in order to be regarded
>as such. And this applies to UML, if its expectations include, as you say,
>not to be just a graphical notation.
>
>Anyway, Dirk, in case you felt my post was rude or offending, I  apologize
>for that: don't feel attacked. I was only trying to expose my point of
>view, maybe I did it with too much strength ;)
>Friendly,
>
>Andrea
>
>
>On Mon, 13 Feb 2006 12:46:15 +0100, Dirk Riehle <[hidden email]> wrote:
>
> >
> >
> >> UML has no attached DL, and it has no precise specification at all of
> >> the
> >> meaning of its elements. It is only intended to defines a graphical
> >> notation and nothing more. Actually, UML has 9 kinds of diagrams which
> >> are
> >
> > UML certainly has higher aspirations than just being a graphical
> > notation. For one, it certainly wants to be a fully executable
> > language one day and people are working on it.
> >
> > There is no reason to leave a major language specified unambiguously
> > deliberately. It just takes time, like any large specification effort.
> >
> >> This is
> >> often considered to obviously hold in the Knolwedge Representation
> >> community where "class" means "set", but it is not so in the context of
> >> object-oriented programming, where for efficiency reasons it is assumed
> >> that objects are instances of a single class and that the class they
are

> >
> > I think you got it backwards. Despite its heritage in programming
> > languages, UML was designed to be expressive, hence the large number
> > of different modeling techniques it offers.
> >
> > OWL and the like have been designed with computability issues in mind
> > (rather than giving the modeler what he needs). The whole extrinsic
> > vs intrinsic definition of classes discussion shows that nicely. If
> > you want an inferencing engine, you need a set-theory like approach.
> > Since UML users typically don't need that, they have not constrained by
> > this.
> >
> > On the specific issue of an object being an instance of multiple
> > classes: I think direct instance-of is overused, massively. What is
> > missing is more use of a concept like is-role-of, and that would do
> > away with these frequently incomprehensibly deep class hierarchies.
> >
> > Dirk
> >
> >
> >
> > Dirk Riehle, ph: +49 172 184 8755, web: http://www.riehle.org
> > Interested in wiki research? Please see http://www.wikisym.org!
> >
>
> -------------------------------------------------------------------------
> > To unsubscribe go to
http://protege.stanford.edu/community/subscribe.html
> >
>
>
>
>--
>Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
>-------------------------------------------------------------------------
>To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html



-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Crowley, Rebecca
In reply to this post by Paul S Prueitt

>>Question:  but is UML a Descriptive Logic ontology?  

Nope - just a convenient way to do the information modeling, transform it into ISO compliant structured metadata, link it to an ontology. Like I said - I think we are going to be reasoning in the ontology and applying it back to the other representations, because there are going to be limitations in trying to reason within the other formalisms.

>>Question 1: what is the DL based ontology that (you mentioned) is now used
>>to link, (semantically link?) the ISO/IEC 11179 standard to?  Is this not
>>Protege Frames?  Is it UML equipped with logic?  I am unclear about what >>you are talking about.

It is NCI Thesaurus. http://nciterms.nci.nih.gov/NCIBrowser/Dictionary.do  I think it now an OBO terminology.

>>Question 2:  When one is talking about 11170 compliance, what does this mean
>>precisely (to you)? http://isotc.iso.org/livelink/livelink/fetch/2000/2489/Ittf_Home/PubliclyAvailableStandards.htm
>>lists the many ISO standards.

Hmm...guessing we are both talking about 11179 (not 11170). To me it means you have to describe your data using the ISO/IEC metamodel for metadata registry. So you have to have Data Element Concepts and they have to have Object Classes and Properties, etc. etc. Much of this is in Part 3. Some other stuff is scattered throughout the other parts. Since I am most definitely not in the business of developing metadata registries, you'd probably get a better answer from somebody who has implemented this standard. I think each of the ISO 11179 registries that I have seen has it's own implementation of the standard. Each one has a slightly different flavor. NCI posts their metamodel on the caDSR site

>>You mention an almost complete paper describing "semantic linkage".  And by
>>"semantic linkage" you mean precisely a mapping from the 11179 to a DL based
>>logic (say Protege Frames (CLIPS) or OWL).

>>As you might know, I (and others) would express some concern about the use
>>of that language, as the real task is structural interoperability between
>>finite state machines.  Right?

I think 'semantic linkage' is the perfect phrase personally. Not sure why that is concerning. It seems to capture exactly what we are doing. As to the 'real task' - I think there are many 'real tasks'. If I understand how you are using the term 'structural interoperability' - caBIG would definitely not consider itself limited to structural interoperability as they are attempting both semantic and syntactic interoperability. The methodology being developed has a healthy dose of human intervention - interoperability reviews, annotation reviews by terminologists, etc. Thank goodness.

I can completely believe that their are going to be methods of combining ontologies and information models that will provide more valid inferences across the model than what we are doing in caBIG. But as I mentioned, I think there are definitely tradeoffs here that few of us may truly appreciate until we have created some models, and used them to exchange data and make some inferences. At the very least, one of the wonderful things about the caBIG project is that we are putting it all to work right away with something like 30-40 applications in three scientific domain workspaces on their way towards integrating through caGrid. That's a good way to find out quickly where things fail.

Glad to see that many different groups are attempting this in different ways. As we all report on what we were able to wring out of these approaches, we should have more data to help us evaluate which paths are worth pursuing for which tasks.

 

best,
rebecca


________________________________

From: Paul S Prueitt [mailto:[hidden email]]
Sent: Sun 2/12/2006 11:20 PM
To: Thomas; Crowley, Rebecca; Protege-Discussion
Subject: side technical discussion on 11179 and DL ontology




Rebecca,

First a review of the discussion between you and Thomas so far... and then
some questions and comments.

You said

" I mentioned that
The Cancer Bioinformatics Grid (caBIG)... has a ISO/IEC
111179 compliant registry that links Data Element Concepts, Valid Values,
Representation Terms, and other
administered components to concepts in a DL (Descriptive Logic) based
ontology - NCI Thesaurus. NCI has made some extensions
to the ISO/IEC standard - which I think they have proposed back to the
standards organization for linking to an ontology in this fashion."


On the link you provided

http://ncicb.nci.nih.gov/NCICB/infrastructure/cacore_overview/cadsr

It is mentioned that

The caCORE objects are represented by UML Models..

Question:  but is UML a Descriptive Logic ontology?  I would not be
surprised if DL can be, is, applied to UML elements.  But, as in the fine
vocabulary of the 11179, the vocabulary of UML is not (in my understanding)
set up perfectly to arrange concepts into hierarchical inheritance trees
with logical "axioms" asserting inferential properties such an transitive
and reflective assignments, as well as the class sub-class definitions.  (I
hope I said that ok.)  In fact this translation issue is a big deal right
now. Right?

You mention an almost complete paper describing "semantic linkage".  And by
"semantic linkage" you mean precisely a mapping from the 11179 to a DL based
logic (say Protege Frames (CLIPS) or OWL).

As you might know, I (and others) would express some concern about the use
of that language, as the real task is structural interoperability between
finite state machines.  Right?

Meaning is properly applied to fixed intended interpretations, of course
this is true,  but a deeper connotation for the term "meaning" comes only
with differential interpretation and in this case the DL ontologies have
both practical and theoretical problems with computability, as has been
pointed out by many.  I hope I am being clear and non-controversal in
pointing this out.


You ask Thomas

"if what you are proposing is to actually implement the ISO standard as
a frame system in protege - I think that would be very useful"


Question 1: what is the DL based ontology that (you mentioned) is now used
to link, (semantically link?) the ISO/IEC 11179 standard to?  Is this not
Protege Frames?  Is it UML equiped with logic?  I am unclear about what you
are talking about.

Question 2:  When one is talking about 11170 compliance, what does this mean
precisely (to you)?

http://isotc.iso.org/livelink/livelink/fetch/2000/2489/Ittf_Home/PubliclyAva
ilableStandards.htm

lists the many ISO standards.

the 11170 has six parts (around 400 pages)..

Fortunately there is a lot of repetition in these parts and within those
pages.

In reading through these pages, I was very impressed with the degree to
which (certain types of) data standardization could be achieved within this
standard.  It is comprehensive and clear.

I am working on a "semantic linkage" between OASIS SOA-IM where again a
quite general understanding of a data modeling framework is available.
Service Oriented Architecture Information Model is far simpler that the
11179, but the key issues are almost the same.

One of the Protege forum contributors, Andrea, made a very proper suggestion
(not that I knew what was proper until I read his note)

http://www.bcngroup.org/beadgames/generativeMethodology/166.htm

I was searching for an answer (in fact I was also seraching to ask the
question correctly), when a stranger (but now a friend) just nailed it!!!!


I should say, not that it makes a lot of difference, that Andrea's use of
the term "non-monotonic" is narrow...  (non-monotonic often means not only
does new information override previous assertions of truth but it does so in
a way in which the location of that which is over written is uncertain.)
But this narrow definition of non-monotonic is wonderfully useful in doing
as much as possible to make structural mappings between things UML, 11179,
and SIA-IM like to things that are like OWL and Protege Frames (ie the
Descriptive Logic type ontologies).

I would appreciate general comments and corrections to my opinions, as we
are all trying to do something that is not fully understood.  So I (as we
all will) will make mistakes ..












-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Paul S Prueitt

Rebecca,

I am very thankful for this discussion and your comments.

After reading I reposted to

http://www.bcngroup.org/beadgames/generativeMethodology/175.htm

so that I could format the conversation better.  We also edit some of the
posts to bring forward the main themes.

The BCNGroup Glass Bead Games is an decade old experiment:

http://www.bcngroup.org/site/beadgames/index.html




The work that Andrea and I are discussing, and that I have some
responsibility for as a R&D project, is to map between object oriented ER,
UML type information models and descriptive logic based ontology such as
Protege Frames , RDF, RDFS, of one of the flavors of OWL.

I agree with this movement towards these types of mapping efforts.

What I feel is missing is an understanding of why this is difficult.

I would suggest that it is !! not merely !! a question of doing some work,
in the same manner as work on related activities have been carried out.

I am a bit tired at the moment, so I would like to stop except to make one
point, not the most important one, but one that is dear to me.

Sometimes what one leaves out in the beginning one can add back later on
with great value added.  But if this is not left out, then one is blocked.


This seems the case with the term "semantics".

Meaning MAY be asignable only in the present moment, which would mean that
meaning could not be formalized.  (This is a conjecture of some literatures
in general systems theory.)

When one adds back humans, then one can add semantics.

http://www.ontologystream.com/home2004.htm

This implies, ie the Second School position implies, that "formal semantics"
is much less that "semantics".  Likewise "expressive logics" is much less
than "expression".  "Artificial intelligence" is much less than
"intelligence".  etc.

There is a caveat, that caveat is that in several ways artificial
intelligence cannot be compared linearly with natural intelligence.  The
essence of the referents are simple different.

The more important discussion is regarding the mapping of object oriented
data models and descriptive/expressive logic based ontology.  The ability to
index information using OWL is significant, but there are profound
limitations to OWL or any future derivation of descriptive/expressive logic
based ontology.  (This is an opinion based on a specific literature.)

Given an understanding of this real limitation allows us to leave out
semantics, ie focus on structural data and process interoperability, and
then later on add the human into the loop.  I think you agree, yes?











-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html

Reply | Threaded
Open this post in threaded view
|

Re: side technical discussion on 11179 and DL ontology

Elisa Kendall
In reply to this post by Paul Prueitt
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

Hi All,

Just to put my two cents in on the OMG perspective: a more precise
characterization of the primary issue
with UML from a knowledge representation perspective is that UML has no
underlying model-theoretic
semantics.  UML is highly expressive, moreso in many regards than many
knowledge representation
paradigms, and is widely used in software engineering, not limited to
object-oriented programming.  Its
features are well documented, and the Object Constraint Language (OCL),
which is a component of the
set of specifications that comprise UML in the large, provides
facilities for writing constraints that are
similar to other rule languages in many respects.  Again, there are no
model-theoretic semantics that I am
aware of for OCL, nor are there inference engines or theorem provers
available that implement OCL as
their underlying rule language.  Given sufficient time and funding,
however, the task of developing such
semantics for OCL and implementing inference capabilities that reflect
it, with some extensions to provide
better support for quantification over certain kinds of sentences seems
doable.

UML's roots, of course, are in object-oriented programming, and thus
while UML instance models can
support multiple inheritence, the subset of the language used to create
metamodels, called the Meta Object
Facility, or MOF, has limitations in this regard that we have recently
been working to overcome (or at
least, deal with), in the Ontology Definition Metamodel (ODM).

The ODM makes no attempt to create model-theoretic semantics for any
aspect of UML.  Rather, we
have used MOF to model the abstract syntax of several knowledge
representation paradigms, including
RDF, OWL, Topic Maps, and Common Logic, reusing the model-theoretic
semantics of  RDF, OWL, and
CL.  This allows us to leverage not only the graphical notation of UML
that is familiar to many, but to
take advantage of MOF's infrastructure and syntactic transformation
capabilities to enable use of
existing intellectual property, such as UML, ER, XML Schema, and other
structured and semi-structured
models as starting points for ontology development.  This is not to say
that such models make good
ontologies, but that the ability to take them as starting points is
often better than nothing, particularly
when attempting to map the semantics of individual resources to a
metadata registries for the purposes
of facilitating interoperability.

The ODM will also provide guidance in transformations from one of these
metamodels to another,
represented in the MOF Query View Transformation (QVT) syntax, which
supports representation and
maintenance of any constructs that may be lost due to lossiness in the
transformations themselves.  In
other words, this work, which is highly visible with and beyond the OMG,
hopes to address some of
what has been described as "semantic linkage" on this thread.  Our work
is being done synergistically
with the W3C Semantic Web Best Practices and Deployment WG, OMG's
Ontology PSIG, and ISO JTC1
working groups responsible for ISO Topic Maps, ISO Common Logic, and the
metadata standards under
discussion here.  The latest draft specification can be found at
http://www.omg.org/cgi-bin/doc?ad/05-09-08.
Another revision, which will include very few metamodel changes, but
additional mapping details and
some revisions to the profile for RDF and OWL will be posted by 3 April
2006 for review.

ISO 11179 was originally intended as a standard for designing and
representing the kinds of constructs
required in metadata registries to support terminology
interoperability.  It is not a modeling paradigm,
such as UML, RDF, or OWL, but can be described by any of them to support
terminology work.
Recent work by Lawrence Berkeley Laboratories, NCI, Mayo Clinic, EPA,
the DoD and others to revise
the 11179 standard (including some of the folks Rebecca mentioned in an
earlier email) has considered
augmenting these capabilities to support registries for model
interoperability, including ontology
interoperability.  While I believe this may be overreaching within a
single standard, particularly without
well defined and broadly reviewed use cases, there is certainly a need
to address model interchange.
A separate ISO effort, 19763, emerging primarily from China, Japan, and
Korea, is aimed at doing so,
including ontology registration. Additionally, there are a number of
less ambitious standards for registry
definition -- UDDI and discussions regarding registration of semantics
for web services (e.g., OWL-S,
SWSL, WSMO, etc.), GILS for information resources, and so forth.  The US
Department of Defense
is currently funding a pilot project to create a series of such
registries -- for semantic web services, for
the semantics of business processes, semantics of information resources,
and more fine-grained
terminological semantics, as a part of a semantically enabled service
oriented architecture
infrastructure.  The first prototype application is for search and
rescue.  Recent presentations on
the ODM and related standards as well as on this SSOA project can be
found at:
http://colab.cim3.net/cgi-bin/wiki.pl?FourthSemanticInteroperabilityforEGovernmentConference_2006_2_0910.
See Friday afternoon session, 1 - 2:30 pm, moderated by Rick Murphy, GSA.

I hope this is helpful --

Elisa



Paul Prueitt wrote:

>Andrea and Dirk,
>
>
>Excellent discussion.
>
>I suggest that the
>
>"The capture of semantics" has a meaning at two levels, and in two very
>different contexts.
>
>It could mean "some" annotation between UML or a object oriented data model
>and some model of the meaning of data/information elements.  In this sense
>it is proper to say, as Rebecca was saying, that a semantic linkage might be
>created as a bridge between a data model and an information model.
>
>It is also proper to say, as some suggest, that DL-based logic and ontology
>specifications can be used to index data that is organized by registries
>like those that are compliant with ISO 11179.  Said in a slightly different
>way, the BioPAX OWL (Full, I believe) rendering of a information model for
>integrating all gene and cell signal pathway data (from very diverse
>bioinformatics projects) my be able to index all existing bioinformatics
>data sources using a DL-type (class - subclass) ontology.  (As I have
>pointed out, this has a huge benefit but some surprises.)
>
>Most know my position regarding the second context.  There are some
>consequences to assuming that the folks at OMG (Object Management Group)
>will one day be able to make UML "expressive" in the sense defined by
>DLbased ontology.
>
>Expressive of "what"?  Two cases exist.
>
>If the data space is stable, well defined with no ambiguity and not complex
>than we all feel that translations between ISO 11179 or SAO IM or other data
>object standard, and DL-type ontology will be figured out this year!!!.  But
>this only means that a translation between two theories related to data
>organization will then exist.  John Sowa's often talked about "lattice of
>theories" comes to mind.
>
>Andrea's solution to the translation of a subset of the SOA IM
>
>http://www.bcngroup.org/beadgames/generativeMethodology/163.htm
>
>is being widely discussed in circles (outside these forums), and is widely
>seen as being "correct".
>
>
>But what about "web service" discovery, where the discovery effort is being
>made in a dynamic environment having lots of novelty and is time
>constrained.
>
>Scenario:  I (a federal agency) am trying to discovery a set of services
>related to how timely response to a natural disaster can be made.  "I" am a
>EU country so these service discovery requests have to be contextualized not
>only into different human languages but with cultural contexts that vary
>even if there is an absolutely agreed to and non ambiguous interlingua
>representation of both relevant DL ontologies and data object information
>models.
>
>
>
>
>
>
>-----Original Message-----
>From: [hidden email]
>[mailto:[hidden email]]On Behalf Of Dirk
>Riehle
>Sent: Monday, February 13, 2006 9:09 AM
>To: [hidden email]
>Subject: [protege-discussion] Re: side technical discussion on 11179 and
>DL ontology
>
>
>
>Hi Andrea,
>
>I'd be surprised to hear anyone at the OMG ever said that UML was
>deliberately designed to be imprecise.  The OMG is driven to a large
>extent by the tool vendors, and lack of interoperability hinders
>their business, so they will eventually fix it. Academic groups, for
>one, have long worked on a formalization of the UML core. I guess it
>will remain an open issue whether it really needs denotational semantics :-)
>
>I work with is-role-of relationships in addition to instance-of
>relationship because it lets me more precisely capture the semantics
>of the domain I'm working on. It really doesn't have much to do with
>the traditional database notion of type migration or dynamically
>attaching objects. Please note that I'm not doing away with
>instance-of (how could I), I'm just saying it is overused.
>
>Take a typical example. If I model a Party (or Person) concept in the
>banking domain, I could have a large class graph with various shades
>of Customer, Debtor, Guarantor, Investor, classes etc. Or I could
>have a Person class in the middle and define concepts of Customer,
>Debtor, etc as roles of a Person around it. I'd argue that roles
>enrich my modeling vocabulary substantially (to the extent that I
>believe it should be part of modeling languages) and let me represent
>a domain more effectively.
>
>I wasn't offended by our original post. I apologize if I came across
>like that---my English is sometimes more curt than I wished it would be.
>
>Dirk
>
>
>
>At 13.02.2006, you wrote:
>
>  
>
>>Hi Dirk,
>>UML was intentionally designed without an unambiguous formal semantics,
>>and after 8 years it still does not have one: indeed, UML is defined by
>>means of itself, i.e. it is specified by its metamodel, which in turn is
>>expressed in UML. This reflect a precise conscious design decision not to
>>give it a formal semantics, and I did not say that this is necessarily
>>bad, but just that it is a strong point of criticism; anyway, I am really
>>glad to hear that someone is eventually working on that. As long as no
>>formal, standard, and agreed-upon specification of its semantics is given,
>>however, UML remains a mere graphical notation. As to the "expressiveness"
>>issue, no quantification of "expressiveness" is possible unless a formal
>>framework that denotes *what* exactly a UML model expresses, and how
>>relationships, classes, and objects exactly relate to one another, is
>>provided.
>>
>>The reason why you perceive that the instance-of relationship is overused
>>and you suggest to replace it with is-role-of properties, in my opinion,
>>is not that you don't really need the semantics of the instance-of
>>relationship, but that you want instance-of relationships to behave like
>>is-role-of properties in ordinary programs, i.e. being able to change
>>dynamically and possibly assume several values. This is probably due to
>>the fact that you are deeply involved in the Software Engineering
>>discipline, and I can understand your point of view because I also was a
>>few time ago. But still, replacing instance-of with is-role-of properties
>>is just a trick, an implementation issue that arises because you have
>>already made a committment to the execution platform (the logical model)
>>in which you are going to translate your conceptual models: that of an
>>object-oriented programming language. Actually, you don't want to give up
>>the semantics of instance-of, you just want to overcome the limitations
>>which are imposed by those environments because they designed to maximize
>>efficiency.
>>
>>Please note, that I am not claiming that object-oriented programming
>>languages are something "improper", just that they have very specific
>>goals, and "expressive" conceptual modeling languages should have a
>>denotational, set-based, model-theoretic semantics in order to be regarded
>>as such. And this applies to UML, if its expectations include, as you say,
>>not to be just a graphical notation.
>>
>>Anyway, Dirk, in case you felt my post was rude or offending, I  apologize
>>for that: don't feel attacked. I was only trying to expose my point of
>>view, maybe I did it with too much strength ;)
>>Friendly,
>>
>>Andrea
>>
>>
>>On Mon, 13 Feb 2006 12:46:15 +0100, Dirk Riehle <[hidden email]> wrote:
>>
>>    
>>
>>>      
>>>
>>>>UML has no attached DL, and it has no precise specification at all of
>>>>the
>>>>meaning of its elements. It is only intended to defines a graphical
>>>>notation and nothing more. Actually, UML has 9 kinds of diagrams which
>>>>are
>>>>        
>>>>
>>>UML certainly has higher aspirations than just being a graphical
>>>notation. For one, it certainly wants to be a fully executable
>>>language one day and people are working on it.
>>>
>>>There is no reason to leave a major language specified unambiguously
>>>deliberately. It just takes time, like any large specification effort.
>>>
>>>      
>>>
>>>>This is
>>>>often considered to obviously hold in the Knolwedge Representation
>>>>community where "class" means "set", but it is not so in the context of
>>>>object-oriented programming, where for efficiency reasons it is assumed
>>>>that objects are instances of a single class and that the class they
>>>>        
>>>>
>are
>  
>
>>>I think you got it backwards. Despite its heritage in programming
>>>languages, UML was designed to be expressive, hence the large number
>>>of different modeling techniques it offers.
>>>
>>>OWL and the like have been designed with computability issues in mind
>>>(rather than giving the modeler what he needs). The whole extrinsic
>>>vs intrinsic definition of classes discussion shows that nicely. If
>>>you want an inferencing engine, you need a set-theory like approach.
>>>Since UML users typically don't need that, they have not constrained by
>>>this.
>>>
>>>On the specific issue of an object being an instance of multiple
>>>classes: I think direct instance-of is overused, massively. What is
>>>missing is more use of a concept like is-role-of, and that would do
>>>away with these frequently incomprehensibly deep class hierarchies.
>>>
>>>Dirk
>>>
>>>
>>>
>>>Dirk Riehle, ph: +49 172 184 8755, web: http://www.riehle.org
>>>Interested in wiki research? Please see http://www.wikisym.org!
>>>
>>>      
>>>
>>-------------------------------------------------------------------------
>>    
>>
>>>To unsubscribe go to
>>>      
>>>
>http://protege.stanford.edu/community/subscribe.html
>  
>
>>
>>--
>>Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
>>-------------------------------------------------------------------------
>>To unsubscribe go to http://protege.stanford.edu/community/subscribe.html
>>    
>>
>
>-------------------------------------------------------------------------
>To unsubscribe go to http://protege.stanford.edu/community/subscribe.html
>
>
>
>-------------------------------------------------------------------------
>To unsubscribe go to http://protege.stanford.edu/community/subscribe.html
>
>
>  
>



-- Attached file removed by Ecartis and put at URL below --
-- Type: text/x-vcard
-- Size: 328 bytes
-- URL : http://protege.stanford.edu/mail_archive/attachments/03-ekendall.vcf


-------------------------------------------------------------------------
To unsubscribe go to http://protege.stanford.edu/community/subscribe.html