RE: Wordnet TF

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

RE: Wordnet TF

John McClure

Hello,
Though WN is a deeply substantive resource, we ultimately determined that WN's
synsets were too informal to be a basis for a commercial ontology. At the same
time, Legal XHTML is designing its ontology based on WN entries. For instance,
events defined by LegalXHTML are classified as either endurant or perdurant
events (acts v activities) -- we look to WN for definitions of both acts and
activities and for hints about their relations and constraints.

Acts and activities fit into a larger economic model. An economy is composed of
sectors, sub-sectors, industry groups, industries, and firms. Firms are composed
of business processes, as required for the manufacture and delivery of goods and
services. Business processes are composed of activities, which are composed of
acts.

With this in mind, a careful design using WN terms bridges between the qualities
it defines (e.g., Endowable), past-participles it defines (e.g., Endowed), acts
that it defines (e.g., EndowmentAct) and processes that it defines (e.g.,
EndowmentActivity). This design ensures, for example, that whenever some
property of EndowmentAct is associated with an Endowable, then incidentally one
can determine that the Endowable is indeed Endowed. By distinguishing an
activity whose 'calendar' includes an EndowmentAct (in addition to other
activities and actions), from the specific act of Endowment, one can identify
those Endowables that may be Endowed from Endowables that are Endowed, and from
Endowables that are not Endowed. Further, it should be equally incidental that
all Endowed resources are, by definition, Endowable resources.

Unfortunately, WN includes neither economic models nor endurant and perdurant
meanings of its terms, critical problems that cannot be overcome by its
translation to OWL. In short, while Wordnet provides a comprehensive linguistic
view of its vocabulary terms, it is unsuitable for rigorous commercial
applications.

John McClure

>-----Original Message-----
>From: [hidden email]
>[mailto:[hidden email]]On Behalf Of Jeremy Carroll
>Sent: Thursday, October 20, 2005 9:27 AM
>To: [hidden email]
>Subject: Wordnet TF
>
>
>
>
>After having expressed a personal lack of enthusiasm for the Wordnet
>Task Force at Monday's telecon, I have discussed this situation with
>colleagues.
>
>The task force has broadly the following jobs to do in phase 1:
>
>   1. define a mapping relating Wordnet and Owl and possibly SKOS
>   2. build consensus for that mapping amongst the several groups who
>have built their own
>   3. build support with the wordnet authors to adopt, distribute and
>maintain the mapping
>   4. possibly develop and deploy a service making Wordnet in RDF
>available on the web
>
>We believe that the Wordnet in RDF/Owl is important and would like to
>ensure that it is completed successfully, but we are aware that time for
>the WG is running out.  We suggest that we should take a realistic view
>of what can be accomplished within the lifetime of the WG.  Then two
>questions arise:
>
>   - is that a useful standalone contribution
>   - how do we get the work completed after SWBP shuts down
>
>
>
>A further point that came up in our discussion is some recent work at
>the Univ. of Chile, inspired by some of the early drafts of the Wordnet TF:
>
>http://www.dcc.uchile.cl/~agraves/wordnet/
>
>This is an additional group that figures under point 2 above.
>
>Jeremy
>
>[[Msg copied by bcc to agraves at dcc.uchile.cl to avoid adding to
>Alvaro's spam]]
>
>
>


Reply | Threaded
Open this post in threaded view
|

RE: Wordnet TF

Aldo Gangemi

Hi John,

comments inside

At 15:39 -0700 28-10-2005, John McClure wrote:
>Hello,
>Though WN is a deeply substantive resource, we ultimately determined that WN's
>synsets were too informal to be a basis for a commercial ontology.

Correct, synsets are lexical entries, not logical classes. In order
to use it as a full-fledged ontology, WN needs a non-trivial
transform (see TF site for documentation).

>At the same
>time, Legal XHTML is designing its ontology based on WN entries. For instance,
>events defined by LegalXHTML are classified as either endurant or perdurant
>events (acts v activities) -- we look to WN for definitions of both acts and
>activities and for hints about their relations and constraints.

Could you please explain a bit more the difference between "endurant
event" and "perdurant event"?
Also the difference between act and activity is not granted. WN takes
activity as a hyponym of act, but most of their senses do not
coincide.

>Acts and activities fit into a larger economic model. An economy is
>composed of
>sectors, sub-sectors, industry groups, industries, and firms. Firms
>are composed
>of business processes, as required for the manufacture and delivery
>of goods and
>services. Business processes are composed of activities, which are composed of
>acts.

Ah ok, you want to define a partonomy of events within an overall
economy (the maximal whole encompassing all economically-relevant
acts, right?).

>With this in mind, a careful design using WN terms bridges between
>the qualities
>it defines (e.g., Endowable), past-participles it defines (e.g.,
>Endowed), acts
>that it defines (e.g., EndowmentAct) and processes that it defines (e.g.,
>EndowmentActivity). This design ensures, for example, that whenever some
>property of EndowmentAct is associated with an Endowable, then
>incidentally one
>can determine that the Endowable is indeed Endowed. By distinguishing an
>activity whose 'calendar' includes an EndowmentAct (in addition to other
>activities and actions), from the specific act of Endowment, one can identify
>those Endowables that may be Endowed from Endowables that are
>Endowed, and from
>Endowables that are not Endowed. Further, it should be equally incidental that
>all Endowed resources are, by definition, Endowable resources.
>
>Unfortunately, WN includes neither economic models nor endurant and perdurant
>meanings of its terms, critical problems that cannot be overcome by its
>translation to OWL. In short, while Wordnet provides a comprehensive
>linguistic
>view of its vocabulary terms, it is unsuitable for rigorous commercial
>applications.

WN is a general-purpose resource. There exist domains that are
assigned to word senses, including some that are economy-oriented.
But even with those, WN is not ideally suited to any domain in
particular, if a detailed ontology is required. WN can be used for
general-purpose projects, as well as for drafting a preliminary
vocabulary for a domain.
Concerning endurant and perdurant, they are usually (e.g in DOLCE,
http://dolce-semanticweb.org) assumed with the approximate meaning of
"object" (endurant) and "process" or "event" (perdurant). You seem to
intend something else. If WN uses your criteria for hyponymy, or can
be transformed to fit your criteria, is not something you can
reasonably expect :). On the other hand, I'm convinced that, once you
clarify the intended meaning of your criteria, it should be easier to
reuse parts of WN for your ontology, e.g. by using OntoWordNet, which
is an OWL transform of WN, linked to DOLCE.

Thanks for interest
Aldo


>John McClure
>
>>-----Original Message-----
>>From: [hidden email]
>>[mailto:[hidden email]]On Behalf Of Jeremy Carroll
>>Sent: Thursday, October 20, 2005 9:27 AM
>>To: [hidden email]
>>Subject: Wordnet TF
>>
>>
>>
>>
>>After having expressed a personal lack of enthusiasm for the Wordnet
>>Task Force at Monday's telecon, I have discussed this situation with
>  >colleagues.
>>
>>The task force has broadly the following jobs to do in phase 1:
>>
>>    1. define a mapping relating Wordnet and Owl and possibly SKOS
>>    2. build consensus for that mapping amongst the several groups who
>>have built their own
>>    3. build support with the wordnet authors to adopt, distribute and
>>maintain the mapping
>>    4. possibly develop and deploy a service making Wordnet in RDF
>>available on the web
>>
>>We believe that the Wordnet in RDF/Owl is important and would like to
>>ensure that it is completed successfully, but we are aware that time for
>>the WG is running out.  We suggest that we should take a realistic view
>>of what can be accomplished within the lifetime of the WG.  Then two
>>questions arise:
>>
>>    - is that a useful standalone contribution
>>    - how do we get the work completed after SWBP shuts down
>>
>>
>>
>>A further point that came up in our discussion is some recent work at
>>the Univ. of Chile, inspired by some of the early drafts of the Wordnet TF:
>>
>>http://www.dcc.uchile.cl/~agraves/wordnet/
>>
>>This is an additional group that figures under point 2 above.
>>
>>Jeremy
>>
>>[[Msg copied by bcc to agraves at dcc.uchile.cl to avoid adding to
>>Alvaro's spam]]
>>
>>
>>


--



Aldo Gangemi
Research Scientist
Laboratory for Applied Ontology
Institute for Cognitive Sciences and Technology
National Research Council (ISTC-CNR)
Via Nomentana 56, 00161, Roma, Italy
Tel: +390644161535
Fax: +390644161513
[hidden email]
http://www.istc.cnr.it/createhtml.php?nbr=71

Reply | Threaded
Open this post in threaded view
|

Re: Wordnet TF

Aldo Gangemi
In reply to this post by John McClure

Hi Jeremy,

At 17:27 +0100 20-10-2005, Jeremy Carroll wrote:
>After having expressed a personal lack of enthusiasm for the Wordnet
>Task Force at Monday's telecon, I have discussed this situation with
>colleagues.
>
>The task force has broadly the following jobs to do in phase 1:
>
>   1. define a mapping relating Wordnet and Owl and possibly SKOS

done for the WN datamodel (see recent development finalized by Mark van Assem)
re:SKOS, I hope to propose something at the F2F
I'd add that we are in touch with ISO people for LMF (Lexical Markup
Framework), and I think it's an advantage to align our WN datamodel
to LMF.

>   2. build consensus for that mapping amongst the several groups who
>have built their own

this is much more related to groups' volunteering ... I'm not going
to provide a detailed analysis of differences in the short term, and
that's possibly a waste of time ... the best our TF can do with the
time remaining is to enforce the OWL datamodel, and ask those several
groups to remark the difference of their datamodels, and we can
eventually tale into account contributions that are more precise, or
that can solve problems we can't solve

>   3. build support with the wordnet authors to adopt, distribute and
>maintain the mapping

we are in touch with Christiane as you probably know, but Princeton's
people is even busier than us, then things don't happen quickly ...
once we have the final datamodel and the web service, things might go
faster anyway

>   4. possibly develop and deploy a service making Wordnet in RDF
>available on the web

this is definitely a task we should accomplish

>
>We believe that the Wordnet in RDF/Owl is important and would like to
>ensure that it is completed successfully, but we are aware that time for
>the WG is running out.  We suggest that we should take a realistic view
>of what can be accomplished within the lifetime of the WG.  Then two
>questions arise:
>
>   - is that a useful standalone contribution

if supported by a service, and with appropriate collaboration with WN
developers and lexical resource standards (e.g. LMF), I think so

>   - how do we get the work completed after SWBP shuts down

the liaison between W3C and ISO that I proposed months ago would be a
good chance to continue the work ... as a co-chair of next LREC 2006
(Linguistic Resources and Evaluation, May, 2006, Genova, Italy) I'm
going to propose a panel on wordnets and the Semantic Web

>
>
>A further point that came up in our discussion is some recent work
>at the Univ. of Chile, inspired by some of the early drafts of the
>Wordnet TF:
>
>http://www.dcc.uchile.cl/~agraves/wordnet/
>
>This is an additional group that figures under point 2 above.

see above; btw Alvaro already contacted me, and if anyone there
volunteers to revise the datamodel, propose an extension, or to
provide a service, that's great

cheers
aldo
--



Aldo Gangemi
Research Scientist
Laboratory for Applied Ontology
Institute for Cognitive Sciences and Technology
National Research Council (ISTC-CNR)
Via Nomentana 56, 00161, Roma, Italy
Tel: +390644161535
Fax: +390644161513
[hidden email]
http://www.istc.cnr.it/createhtml.php?nbr=71

Reply | Threaded
Open this post in threaded view
|

[WN] Endurant Objects?

John McClure
In reply to this post by Aldo Gangemi

Hi Aldo,
>Concerning endurant and perdurant, they are usually (e.g in DOLCE,
>http://dolce-semanticweb.org) assumed with the approximate meaning of
>"object" (endurant) and "process" or "event" (perdurant).

Hmm, my understanding is that endurant and perdurant are descriptors applicable
to qualities and quantities, that is, to *attributes of* resources. A resource
has endurant and/or perdurant attributes. For example, a person's height is a
perdurant quantity; a person's eye color is an endurant quality.

Apart from that, events-as-process track to my view also of perdurants. However,
there is a huge difference between perdurant events and endurant events. (Note:
I consider an event to be an attribute of the thing to which it occurs, as it
relates to the 'state' of the resource.) A marriage ceremony is a process; a
recital of a marriage vow is a one-shot action consisting of no process
whatsoever. Maybe what needs to be done is to further define what a 'process'
is, from the DOLCE view? My own view is that a process is a series of actions
and subprocesses. An action (which normally occurs in the context of a process)
has no 'sub-action' and of course no subprocess component.

So, I am concerned about the presumption of a parallel between 'endurant' and
'object' -- I think it's somewhat misleading if not wrong. As for the larger
picture, sure, I agree that hooking a controlled, commercialized subset of WN
back to a transformed WN-ontology can be a good idea, although I'm not sure
about doing so until the subset is stable.

BTW, for a view of (the heart of) the Legal XHTML model -- where events are
prime (by your actions are ye known!) -- please see
http://www.hypergrove.com/OWL/ and navigate to 1) Resource Model and 2) Event
Model Our model allows "persons, places, and things" defined by other ontologies
to be 'pluggable' into the LegalXHTML ontology. For instance, a "Person" in
another model could subclass LegalXHTML's "ContactableThing" in order to pick up
Contact-related attributes. LegalXHTML aims to package its attributes, as much
as possible, as perdurant and endurant events. The model published (today) is a
first big step towards that goal.

Thanks for your reply,
John McClure


Reply | Threaded
Open this post in threaded view
|

Re: [WN] Endurant Objects?

Aldo Gangemi

Hi John,

At 12:11 -0800 31-10-2005, John McClure wrote:

>Hi Aldo,
>>Concerning endurant and perdurant, they are usually (e.g in DOLCE,
>>http://dolce-semanticweb.org) assumed with the approximate meaning of
>>"object" (endurant) and "process" or "event" (perdurant).
>
>Hmm, my understanding is that endurant and perdurant are descriptors
>applicable
>to qualities and quantities, that is, to *attributes of* resources. A resource
>has endurant and/or perdurant attributes. For example, a person's height is a
>perdurant quantity; a person's eye color is an endurant quality.

Strictly speaking, these examples depend on context:  height changes
significantly for only part of a person lifecycle, while even eye
color can change in later phases, let alone in newborns. But this is
just for general talking.

>Apart from that, events-as-process track to my view also of
>perdurants. However,
>there is a huge difference between perdurant events and endurant
>events. (Note:
>I consider an event to be an attribute of the thing to which it occurs, as it
>relates to the 'state' of the resource.)

This use of state and event is a bit peculiar. In system theory (but
also in engineering, Petri Nets, etc.), it's often assumed that a
process is constituted of states whose boundaries are events that
occur (transitions).
In so-called 4-dimensionalist ontologies, events are attributes in
the sense they are temporal parts of an entity.
BTW, I still miss what it means for events to "endure" or to
"perdure" ... (after looking at your site) ... ok, I think you catch
a know difference related to "aspects" in event structure, by which
we can consider an event as a whole (your "endurant event"), or as a
composition of parts (your "perdurant event"). But aspects are not
inherent in the processes you're trying to describe, they come from
the perspective of the user. E.g., if I went to an auction and got a
lamp, I can see the auction as a single, non-analyzed whole that
provided me a lamp, or as a long series of events that allowed me to
buy a lamp. Therefore, if I'm just interested in the fact that I've
got a lamp, that distinction is totally irrelevant, while if I'm
interested in legal consequences, I should assume the second
perspective.

>A marriage ceremony is a process; a
>recital of a marriage vow is a one-shot action consisting of no process
>whatsoever. Maybe what needs to be done is to further define what a 'process'
>is, from the DOLCE view?

You seem to assume that an action is atomic, while processes have
parts. So far so good. I'm only trying to catch the intended meaning
of your distinctions.

>My own view is that a process is a series of actions
>and subprocesses. An action (which normally occurs in the context of
>a process)
>has no 'sub-action' and of course no subprocess component.

Consequently to what you said, indeed.

>So, I am concerned about the presumption of a parallel between 'endurant' and
>'object' -- I think it's somewhat misleading if not wrong.

Why concerned? I was trying to compare your use of terms to another one.
There is nothing "wrong" in using terms in a controlled way, if
appropriate axioms or explanations prevent misunderstanding. For
example, DOLCE has hundreds of axioms and a lot of documentation that
attempt at clarifying the intended meaning of terms in its logical
vocabulary.
Of course, if you try to convince anyone to use "cat" in order to
mean "dog", you may have hard time in talking to people, but this is
another story.

>As for the larger
>picture, sure, I agree that hooking a controlled, commercialized subset of WN
>back to a transformed WN-ontology can be a good idea, although I'm not sure
>about doing so until the subset is stable.
>
>BTW, for a view of (the heart of) the Legal XHTML model -- where events are
>prime (by your actions are ye known!) -- please see
>http://www.hypergrove.com/OWL/ and navigate to 1) Resource Model and 2) Event
>Model

Interesting. Specially your design pattern for states and qualities.
Your use of rdf typing to relate conceptual and natural language
syntactic information is creative, while I don't see the point in
using metaclasses for end/perd events.
For other approaches and patterns to represent legal reality see also
the recent "Law and the Semantic Web", R. Benjamins et al., Springer.
Other comments when loading the owl file into some editor (the Event
file is still loading on Protege after 30 minutes).

>  Our model allows "persons, places, and things" defined by other ontologies
>to be 'pluggable' into the LegalXHTML ontology. For instance, a "Person" in
>another model could subclass LegalXHTML's "ContactableThing" in
>order to pick up
>Contact-related attributes.

As it is the case for any ontology with some generality ;)

>LegalXHTML aims to package its attributes, as much
>as possible, as perdurant and endurant events. The model published
>(today) is a
>first big step towards that goal.
>

More after loading
Cheers
Aldo
--



Aldo Gangemi
Research Scientist
Laboratory for Applied Ontology
Institute for Cognitive Sciences and Technology
National Research Council (ISTC-CNR)
Via Nomentana 56, 00161, Roma, Italy
Tel: +390644161535
Fax: +390644161513
[hidden email]
http://www.istc.cnr.it/createhtml.php?nbr=71

Reply | Threaded
Open this post in threaded view
|

Re: [WN] Endurant Objects?

Alan Rector

Aldo, John

I am coming in late on this, but I am a bit surprised.

I thought Perdurant/Occurrent and Endurant/Continuant were of sortals  
rather than qualities, where the problems are clearly different.    
This seems an easy distinction to make.  In my biomedical world very  
few qaualities of anything are truly 'endurant' - i.e. unchangeable.

Likewise, I've always thought that the event/process distinction was  
one of granularity/perspective.  If our temporal mesh is fine enough  
all events above the subatomic can be redefined as a series of  
processes.  That's certainly the view of standard texts on the topic  
such as Davies.

Alan




On 1 Nov 2005, at 22:49, Aldo Gangemi wrote:

>
> Hi John,
>
> At 12:11 -0800 31-10-2005, John McClure wrote:
>
>> Hi Aldo,
>>
>>> Concerning endurant and perdurant, they are usually (e.g in DOLCE,
>>> http://dolce-semanticweb.org) assumed with the approximate  
>>> meaning of
>>> "object" (endurant) and "process" or "event" (perdurant).
>>>
>>
>> Hmm, my understanding is that endurant and perdurant are  
>> descriptors applicable
>> to qualities and quantities, that is, to *attributes of*  
>> resources. A resource
>> has endurant and/or perdurant attributes. For example, a person's  
>> height is a
>> perdurant quantity; a person's eye color is an endurant quality.
>>
>
> Strictly speaking, these examples depend on context:  height  
> changes significantly for only part of a person lifecycle, while  
> even eye color can change in later phases, let alone in newborns.  
> But this is just for general talking.
>
>
>> Apart from that, events-as-process track to my view also of  
>> perdurants. However,
>> there is a huge difference between perdurant events and endurant  
>> events. (Note:
>> I consider an event to be an attribute of the thing to which it  
>> occurs, as it
>> relates to the 'state' of the resource.)
>>
>
> This use of state and event is a bit peculiar. In system theory  
> (but also in engineering, Petri Nets, etc.), it's often assumed  
> that a process is constituted of states whose boundaries are events  
> that occur (transitions).
> In so-called 4-dimensionalist ontologies, events are attributes in  
> the sense they are temporal parts of an entity.
> BTW, I still miss what it means for events to "endure" or to  
> "perdure" ... (after looking at your site) ... ok, I think you  
> catch a know difference related to "aspects" in event structure, by  
> which we can consider an event as a whole (your "endurant event"),  
> or as a composition of parts (your "perdurant event"). But aspects  
> are not inherent in the processes you're trying to describe, they  
> come from the perspective of the user. E.g., if I went to an  
> auction and got a lamp, I can see the auction as a single, non-
> analyzed whole that provided me a lamp, or as a long series of  
> events that allowed me to buy a lamp. Therefore, if I'm just  
> interested in the fact that I've got a lamp, that distinction is  
> totally irrelevant, while if I'm interested in legal consequences,  
> I should assume the second perspective.
>
>
>> A marriage ceremony is a process; a
>> recital of a marriage vow is a one-shot action consisting of no  
>> process
>> whatsoever. Maybe what needs to be done is to further define what  
>> a 'process'
>> is, from the DOLCE view?
>>
>
> You seem to assume that an action is atomic, while processes have  
> parts. So far so good. I'm only trying to catch the intended  
> meaning of your distinctions.
>
>
>> My own view is that a process is a series of actions
>> and subprocesses. An action (which normally occurs in the context  
>> of a process)
>> has no 'sub-action' and of course no subprocess component.
>>
>
> Consequently to what you said, indeed.
>
>
>> So, I am concerned about the presumption of a parallel between  
>> 'endurant' and
>> 'object' -- I think it's somewhat misleading if not wrong.
>>
>
> Why concerned? I was trying to compare your use of terms to another  
> one.
> There is nothing "wrong" in using terms in a controlled way, if  
> appropriate axioms or explanations prevent misunderstanding. For  
> example, DOLCE has hundreds of axioms and a lot of documentation  
> that attempt at clarifying the intended meaning of terms in its  
> logical vocabulary.
> Of course, if you try to convince anyone to use "cat" in order to  
> mean "dog", you may have hard time in talking to people, but this  
> is another story.
>
>
>> As for the larger
>> picture, sure, I agree that hooking a controlled, commercialized  
>> subset of WN
>> back to a transformed WN-ontology can be a good idea, although I'm  
>> not sure
>> about doing so until the subset is stable.
>>
>> BTW, for a view of (the heart of) the Legal XHTML model -- where  
>> events are
>> prime (by your actions are ye known!) -- please see
>> http://www.hypergrove.com/OWL/ and navigate to 1) Resource Model  
>> and 2) Event
>> Model
>>
>
> Interesting. Specially your design pattern for states and  
> qualities. Your use of rdf typing to relate conceptual and natural  
> language syntactic information is creative, while I don't see the  
> point in using metaclasses for end/perd events.
> For other approaches and patterns to represent legal reality see  
> also the recent "Law and the Semantic Web", R. Benjamins et al.,  
> Springer.
> Other comments when loading the owl file into some editor (the  
> Event file is still loading on Protege after 30 minutes).
>
>
>>  Our model allows "persons, places, and things" defined by other  
>> ontologies
>> to be 'pluggable' into the LegalXHTML ontology. For instance, a  
>> "Person" in
>> another model could subclass LegalXHTML's "ContactableThing" in  
>> order to pick up
>> Contact-related attributes.
>>
>
> As it is the case for any ontology with some generality ;)
>
>
>> LegalXHTML aims to package its attributes, as much
>> as possible, as perdurant and endurant events. The model published  
>> (today) is a
>> first big step towards that goal.
>>
>>
>
> More after loading
> Cheers
> Aldo
> --
>
>
>
> Aldo Gangemi
> Research Scientist
> Laboratory for Applied Ontology
> Institute for Cognitive Sciences and Technology
> National Research Council (ISTC-CNR)
> Via Nomentana 56, 00161, Roma, Italy
> Tel: +390644161535
> Fax: +390644161513
> [hidden email]
> http://www.istc.cnr.it/createhtml.php?nbr=71
>
>

-----------------------
Alan Rector
Professor of Medical Informatics
Department of Computer Science
University of Manchester
Manchester M13 9PL, UK
TEL +44 (0) 161 275 6188/6149
FAX +44 (0) 161 275 6204
www.cs.man.ac.uk/mig
www.clinical-esciences.org
www.co-ode.org


Reply | Threaded
Open this post in threaded view
|

Re: [WN] Endurant Objects?

Aldo Gangemi

Hi Alan,

At 7:42 +0000 2-11-2005, Alan Rector wrote:
>Aldo, John
>
>I am coming in late on this, but I am a bit surprised.
>
>I thought Perdurant/Occurrent and Endurant/Continuant were of
>sortals rather than qualities, where the problems are clearly
>different.

Totally agreed. Consider btw that 4-dimensionalists (e.g. in trope
theory) are closer to John's intuition (but they do not use, to my
knowlegde, the end/perd terminology).

>
>This seems an easy distinction to make.  In my biomedical world very
>few qaualities of anything are truly 'endurant' - i.e. unchangeable.

As a matter of fact, the relation defined in DOLCE between entities
and qualities, called there "inherence", is temporalized by default
(such temporalization requires an n-ary pattern to be expressed in
OWL btw).

>Likewise, I've always thought that the event/process distinction was
>one of granularity/perspective.  If our temporal mesh is fine enough
>all events above the subatomic can be redefined as a series of
>processes.  That's certainly the view of standard texts on the topic
>such as Davies.

Totally agreed again. This is also the common view from aspectual
semantics in linguistics (e.g. Dowty).

That's good then; nevertheless, concerning John's issues, it's
important to separate *terminological* from *conceptual*
distinctions. If LegalXHTML vocabulary wants to use end/perd,
event/state and other terms according to John's intended meaning, no
problem about that, provided that it's possible to interpret it on
formal grounds.
Consider that LegalXHTML is in OWL-Full, defines a kind of type
theory, some metaclasses are for grammatical categories, and
redefines notions like subClass, ISA, etc. This approach complicates
the discussion, because not only have we to negotiate the intended
meaning of classes and properties, but also of their formal semantics.

Ciao
Aldo

>Alan
>
>
>
>
>On 1 Nov 2005, at 22:49, Aldo Gangemi wrote:
>
>>
>>Hi John,
>>
>>At 12:11 -0800 31-10-2005, John McClure wrote:
>>
>>>Hi Aldo,
>>>
>>>>Concerning endurant and perdurant, they are usually (e.g in DOLCE,
>>>>http://dolce-semanticweb.org) assumed with the approximate meaning of
>>>>"object" (endurant) and "process" or "event" (perdurant).
>>>>
>>>
>>>Hmm, my understanding is that endurant and perdurant are
>>>descriptors applicable
>>>to qualities and quantities, that is, to *attributes of*
>>>resources. A resource
>>>has endurant and/or perdurant attributes. For example, a person's
>>>height is a
>>>perdurant quantity; a person's eye color is an endurant quality.
>>>
>>
>>Strictly speaking, these examples depend on context:  height
>>changes significantly for only part of a person lifecycle, while
>>even eye color can change in later phases, let alone in newborns.
>>But this is just for general talking.
>>
>>>Apart from that, events-as-process track to my view also of
>>>perdurants. However,
>>>there is a huge difference between perdurant events and endurant
>>>events. (Note:
>>>I consider an event to be an attribute of the thing to which it
>>>occurs, as it
>>>relates to the 'state' of the resource.)
>>>
>>
>>This use of state and event is a bit peculiar. In system theory
>>(but also in engineering, Petri Nets, etc.), it's often assumed
>>that a process is constituted of states whose boundaries are events
>>that occur (transitions).
>>In so-called 4-dimensionalist ontologies, events are attributes in
>>the sense they are temporal parts of an entity.
>>BTW, I still miss what it means for events to "endure" or to
>>"perdure" ... (after looking at your site) ... ok, I think you
>>catch a know difference related to "aspects" in event structure, by
>>which we can consider an event as a whole (your "endurant event"),
>>or as a composition of parts (your "perdurant event"). But aspects
>>are not inherent in the processes you're trying to describe, they
>>come from the perspective of the user. E.g., if I went to an
>>auction and got a lamp, I can see the auction as a single,
>>non-analyzed whole that provided me a lamp, or as a long series of
>>events that allowed me to buy a lamp. Therefore, if I'm just
>>interested in the fact that I've got a lamp, that distinction is
>>totally irrelevant, while if I'm interested in legal consequences,
>>I should assume the second perspective.
>>
>>>A marriage ceremony is a process; a
>>>recital of a marriage vow is a one-shot action consisting of no process
>>>whatsoever. Maybe what needs to be done is to further define what
>>>a 'process'
>>>is, from the DOLCE view?
>>>
>>
>>You seem to assume that an action is atomic, while processes have
>>parts. So far so good. I'm only trying to catch the intended
>>meaning of your distinctions.
>>
>>>My own view is that a process is a series of actions
>>>and subprocesses. An action (which normally occurs in the context
>>>of a process)
>>>has no 'sub-action' and of course no subprocess component.
>>>
>>
>>Consequently to what you said, indeed.
>>
>>>So, I am concerned about the presumption of a parallel between
>>>'endurant' and
>>>'object' -- I think it's somewhat misleading if not wrong.
>>>
>>
>>Why concerned? I was trying to compare your use of terms to another one.
>>There is nothing "wrong" in using terms in a controlled way, if
>>appropriate axioms or explanations prevent misunderstanding. For
>>example, DOLCE has hundreds of axioms and a lot of documentation
>>that attempt at clarifying the intended meaning of terms in its
>>logical vocabulary.
>>Of course, if you try to convince anyone to use "cat" in order to
>>mean "dog", you may have hard time in talking to people, but this
>>is another story.
>>
>>>As for the larger
>>>picture, sure, I agree that hooking a controlled, commercialized
>>>subset of WN
>>>back to a transformed WN-ontology can be a good idea, although I'm not sure
>>>about doing so until the subset is stable.
>>>
>>>BTW, for a view of (the heart of) the Legal XHTML model -- where events are
>>>prime (by your actions are ye known!) -- please see
>>>http://www.hypergrove.com/OWL/ and navigate to 1) Resource Model
>>>and 2) Event
>>>Model
>>>
>>
>>Interesting. Specially your design pattern for states and
>>qualities. Your use of rdf typing to relate conceptual and natural
>>language syntactic information is creative, while I don't see the
>>point in using metaclasses for end/perd events.
>>For other approaches and patterns to represent legal reality see
>>also the recent "Law and the Semantic Web", R. Benjamins et al.,
>>Springer.
>>Other comments when loading the owl file into some editor (the
>>Event file is still loading on Protege after 30 minutes).
>>
>>>  Our model allows "persons, places, and things" defined by other ontologies
>>>to be 'pluggable' into the LegalXHTML ontology. For instance, a "Person" in
>>>another model could subclass LegalXHTML's "ContactableThing" in
>>>order to pick up
>>>Contact-related attributes.
>>>
>>
>>As it is the case for any ontology with some generality ;)
>>
>>>LegalXHTML aims to package its attributes, as much
>>>as possible, as perdurant and endurant events. The model published
>>>(today) is a
>>>first big step towards that goal.
>>>
>>
>>More after loading
>>Cheers
>>Aldo
>>--
>>
>>
>>
>>Aldo Gangemi
>>Research Scientist
>>Laboratory for Applied Ontology
>>Institute for Cognitive Sciences and Technology
>>National Research Council (ISTC-CNR)
>>Via Nomentana 56, 00161, Roma, Italy
>>Tel: +390644161535
>>Fax: +390644161513
>>[hidden email]
>>http://www.istc.cnr.it/createhtml.php?nbr=71
>>
>
>-----------------------
>Alan Rector
>Professor of Medical Informatics
>Department of Computer Science
>University of Manchester
>Manchester M13 9PL, UK
>TEL +44 (0) 161 275 6188/6149
>FAX +44 (0) 161 275 6204
>www.cs.man.ac.uk/mig
>www.clinical-esciences.org
>www.co-ode.org


--



Aldo Gangemi
Research Scientist
Laboratory for Applied Ontology
Institute for Cognitive Sciences and Technology
National Research Council (ISTC-CNR)
Via Nomentana 56, 00161, Roma, Italy
Tel: +390644161535
Fax: +390644161513
[hidden email]
http://www.istc.cnr.it/createhtml.php?nbr=71

Reply | Threaded
Open this post in threaded view
|

Re: [WN] Endurant Objects?

Pat Hayes
In reply to this post by Alan Rector

>Aldo, John
>
>I am coming in late on this, but I am a bit surprised.

Me too. I would strongly suggest that these distinctions are more
harmful than useful, except possibly in artificial settings which
allow them to be defined clearly, and then those locally appropriate
definitions should be used and stated clearly, and preferably the
distinctions made using a vocabulary special to that domain, so as
not to confuse it with related but not identical distinctions made in
other domains.

These distinctions are certainly not philosophically necessary or in
any sense required for clear expression of domain facts; and all such
distinctions are in fact contested by various schools of thought in
philosophical ontology. In a word, never take such basic distinctions
on trust as having been fixed by an external authority.

>I thought Perdurant/Occurrent and Endurant/Continuant were of
>sortals rather than qualities

There is no fact of the matter. They are used in both ways, and some
authors treat them differently from others.

>, where the problems are clearly different.
>This seems an easy distinction to make.  In my biomedical world very
>few qaualities of anything are truly 'endurant' - i.e. unchangeable.

Quite. Part of what makes these categories so suspicious and
potentially harmful is that they are stated as very general,
'upper-level' distinctions, but different domains and worlds of
application will draw the boundaries between the categories
differently, or categorize the same thing in conflicting ways, or
refuse to recognize some categories at all. (Also, different
philosopher's definitions of these terms will often give conflicting
results when applied to cases which the philosopher did not think
of.)  This creates barriers to interoperabilty which are entirely
artificial, and simply go away when these upper-level taxonomies are
rejected, which I would strongly recommend they should be. This
particular occurrent/continuant distinction is almost uniquely
poisonous, has generated libraries full of controversy, and as far as
I can detect has not been the slightest practical use in support of
any actual inferences. (Like all binary taxonomic classifications, it
can be used to help detect gross conceptual errors in large
multi-authored knowledge bases; and it can be related to NL
distinctions between noun and verb, but only superficially.) Its
intellectual roots are in late 19th-century phenomenology: it should
be called the Brentano Mistake.

>Likewise, I've always thought that the event/process distinction was
>one of granularity/perspective.

I agree; and it also can be changed on other grounds also.

>  If our temporal mesh is fine enough all events above the subatomic

and even there, for some purposes.

Pat Hayes

>can be redefined as a series of processes.  That's certainly the
>view of standard texts on the topic such as Davies.
>
>Alan
>
>
>
>
>On 1 Nov 2005, at 22:49, Aldo Gangemi wrote:
>
>>
>>Hi John,
>>
>>At 12:11 -0800 31-10-2005, John McClure wrote:
>>
>>>Hi Aldo,
>>>
>>>>Concerning endurant and perdurant, they are usually (e.g in DOLCE,
>>>>http://dolce-semanticweb.org) assumed with the approximate meaning of
>>>>"object" (endurant) and "process" or "event" (perdurant).
>>>>
>>>
>>>Hmm, my understanding is that endurant and perdurant are
>>>descriptors applicable
>>>to qualities and quantities, that is, to *attributes of*
>>>resources. A resource
>>>has endurant and/or perdurant attributes. For example, a person's
>>>height is a
>>>perdurant quantity; a person's eye color is an endurant quality.
>>>
>>
>>Strictly speaking, these examples depend on context:  height
>>changes significantly for only part of a person lifecycle, while
>>even eye color can change in later phases, let alone in newborns.
>>But this is just for general talking.
>>
>>>Apart from that, events-as-process track to my view also of
>>>perdurants. However,
>>>there is a huge difference between perdurant events and endurant
>>>events. (Note:
>>>I consider an event to be an attribute of the thing to which it
>>>occurs, as it
>>>relates to the 'state' of the resource.)
>>>
>>
>>This use of state and event is a bit peculiar. In system theory
>>(but also in engineering, Petri Nets, etc.), it's often assumed
>>that a process is constituted of states whose boundaries are events
>>that occur (transitions).
>>In so-called 4-dimensionalist ontologies, events are attributes in
>>the sense they are temporal parts of an entity.
>>BTW, I still miss what it means for events to "endure" or to
>>"perdure" ... (after looking at your site) ... ok, I think you
>>catch a know difference related to "aspects" in event structure, by
>>which we can consider an event as a whole (your "endurant event"),
>>or as a composition of parts (your "perdurant event"). But aspects
>>are not inherent in the processes you're trying to describe, they
>>come from the perspective of the user. E.g., if I went to an
>>auction and got a lamp, I can see the auction as a single,
>>non-analyzed whole that provided me a lamp, or as a long series of
>>events that allowed me to buy a lamp. Therefore, if I'm just
>>interested in the fact that I've got a lamp, that distinction is
>>totally irrelevant, while if I'm interested in legal consequences,
>>I should assume the second perspective.
>>
>>>A marriage ceremony is a process; a
>>>recital of a marriage vow is a one-shot action consisting of no process
>>>whatsoever. Maybe what needs to be done is to further define what
>>>a 'process'
>>>is, from the DOLCE view?
>>>
>>
>>You seem to assume that an action is atomic, while processes have
>>parts. So far so good. I'm only trying to catch the intended
>>meaning of your distinctions.
>>
>>>My own view is that a process is a series of actions
>>>and subprocesses. An action (which normally occurs in the context
>>>of a process)
>>>has no 'sub-action' and of course no subprocess component.
>>>
>>
>>Consequently to what you said, indeed.
>>
>>>So, I am concerned about the presumption of a parallel between
>>>'endurant' and
>>>'object' -- I think it's somewhat misleading if not wrong.
>>>
>>
>>Why concerned? I was trying to compare your use of terms to another one.
>>There is nothing "wrong" in using terms in a controlled way, if
>>appropriate axioms or explanations prevent misunderstanding. For
>>example, DOLCE has hundreds of axioms and a lot of documentation
>>that attempt at clarifying the intended meaning of terms in its
>>logical vocabulary.
>>Of course, if you try to convince anyone to use "cat" in order to
>>mean "dog", you may have hard time in talking to people, but this
>>is another story.
>>
>>>As for the larger
>>>picture, sure, I agree that hooking a controlled, commercialized
>>>subset of WN
>>>back to a transformed WN-ontology can be a good idea, although I'm not sure
>>>about doing so until the subset is stable.
>>>
>>>BTW, for a view of (the heart of) the Legal XHTML model -- where events are
>>>prime (by your actions are ye known!) -- please see
>>>http://www.hypergrove.com/OWL/ and navigate to 1) Resource Model
>>>and 2) Event
>>>Model
>>>
>>
>>Interesting. Specially your design pattern for states and
>>qualities. Your use of rdf typing to relate conceptual and natural
>>language syntactic information is creative, while I don't see the
>>point in using metaclasses for end/perd events.
>>For other approaches and patterns to represent legal reality see
>>also the recent "Law and the Semantic Web", R. Benjamins et al.,
>>Springer.
>>Other comments when loading the owl file into some editor (the
>>Event file is still loading on Protege after 30 minutes).
>>
>>>  Our model allows "persons, places, and things" defined by other ontologies
>>>to be 'pluggable' into the LegalXHTML ontology. For instance, a "Person" in
>>>another model could subclass LegalXHTML's "ContactableThing" in
>>>order to pick up
>>>Contact-related attributes.
>>>
>>
>>As it is the case for any ontology with some generality ;)
>>
>>>LegalXHTML aims to package its attributes, as much
>>>as possible, as perdurant and endurant events. The model published
>>>(today) is a
>>>first big step towards that goal.
>>>
>>
>>More after loading
>>Cheers
>>Aldo
>>--
>>
>>
>>
>>Aldo Gangemi
>>Research Scientist
>>Laboratory for Applied Ontology
>>Institute for Cognitive Sciences and Technology
>>National Research Council (ISTC-CNR)
>>Via Nomentana 56, 00161, Roma, Italy
>>Tel: +390644161535
>>Fax: +390644161513
>>[hidden email]
>>http://www.istc.cnr.it/createhtml.php?nbr=71
>>
>
>-----------------------
>Alan Rector
>Professor of Medical Informatics
>Department of Computer Science
>University of Manchester
>Manchester M13 9PL, UK
>TEL +44 (0) 161 275 6188/6149
>FAX +44 (0) 161 275 6204
>www.cs.man.ac.uk/mig
>www.clinical-esciences.org
>www.co-ode.org


--
---------------------------------------------------------------------
IHMC (850)434 8903 or (650)494 3973   home
40 South Alcaniz St. (850)202 4416   office
Pensacola (850)202 4440   fax
FL 32502 (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes


Reply | Threaded
Open this post in threaded view
|

RE: [WN] Endurant Objects?

John McClure
In reply to this post by Aldo Gangemi

Folks,
As [1] says, "enduring and perduring entities ... [are] a distinction still
strongly debated both in the philosophical literature and within ontology
standardization initiatives."  My use of these terms as adjectival qualifiers
seems at odds with [2] where is said "an exclusive use of these **nouns** seems
to reflect an exclusive interest in quality-bearers ..." Nevertheless I do feel
that, given a resource is merely a reflection of its qualities and quantities,
then the notion of an enduring or perduring object is therefore derived from the
extent to which its attributes are enduring or perduring.

But is this endurant/perdurant distinction actually useful as I construct the
LegalXHTML ontology? My answer: it depends on if the design of the ontology's
subclass hierarchy distinguishes between them...  DOLCE is a fine example of
such a model... but a better approach for Legal XHTML is to (eventually) type
its classes with metaclasses like PhysicalEndurant, NonAgentiveSocialObject, and
AbstractRegion. In other words, I suggest it most effective to link concrete
classes with these soft constructs by using <rdf:type>, rather than
<rdfs:subClassOf>.

I want my subclass hierarchies as flat as possible so as to easily refactor
later, based on user feedback. The subclass hierarchies need to be as simple as
possible for commercial users (who likely don't care about endurant v
perdurant). Classifying LegalXHTML terms with ontological and grammatical
metaclasses (coupled with associated class annotations pointing at operative
properties related to the metaclasses) seems a strategy that can meet the needs
of skillful reviewers like yourselves without sacrificing the simplicity needed
in commercial environments.

At the same time, please note that LegalXHTML defines the text attribute
<asOf>date</asOf>. This attribute applies to any LegalXHTML resource and to any
LegalXHTML object attribute. This attribute is the device for identifying one
snapshot versus another, of any resource or of any resource attribute, whether
perdurant or endurant. [BTW, I do believe that <asOf> at a minumum represents a
darn basic 'best-practice' itself.]

Aldo said:
>Consider that LegalXHTML is in OWL-Full, defines a kind of type
>theory, some metaclasses are for grammatical categories, and
>redefines notions like subClass, ISA, etc. This approach complicates
>the discussion, because not only have we to negotiate the intended
>meaning of classes and properties, but also of their formal semantics.

I didn't realize that LegalXHTML is redefining notions like subClass, ISA, etc
(where?) and I don't know how LegalXHTML "defines a kind of type theory".
Definitely I appreciate any suggestions how to un-complicate the discussion. My
goal is to be as vanilla (un-controversial) as possible.

>>Interesting. Specially your design pattern for states and
>>qualities. Your use of rdf typing to relate conceptual and natural
>>language syntactic information is creative, while I don't see the
>>point in using metaclasses for end/perd events.

My use for the EndurantEvent and PerdurantEvent metaclasses is to present
separate lists of acts v activities. I'm happy to change from EndurantEvent to
ActionClass and PerdurantEvent to ActivityClass. Incidentally I can't use
'Process' because UNCEFACT already catalogs (business) processes -- one
objective I have is to decompose their processes into activities, and activities
into acts and sub-activities.

Thanks -
John McClure

[1] http://wonderweb.semanticweb.org/deliverables/documents/D18.pdf
[2] http://hem.passagen.se/ijohansson/information2.PDF