[Top] [All Lists]

Re: [ontolog-forum] Ontology Project Organization:

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Mike Bennett <mbennett@xxxxxxxxxxxxxxx>
Date: Mon, 01 Jun 2009 00:14:36 +0100
Message-id: <4A230F5C.2010109@xxxxxxxxxxxxxxx>
Hi Patrick,    (01)

Sorry for the delayed reply.    (02)

A couple of thoughts interleaved below...    (03)

Patrick Cassidy wrote:
>> Independent/Relative/Mediating
>> Concrete/Abstract
>> Continuant/Occurrent
> Mike, I haven't found partitions along those lines to be useful.  There are
> disjoint categories such as "Event" and "Object" and "AttributeValue" to
> which I can give some necessary conditions.  But in trying to represent the
> basic elements of thought, which are reflected in the basic language, there
> are too many ways of viewing things to make it practical to try to squeeze
> all those views into a few top partitions.  
>       (04)

This is precisely why I would not want to try to represent all the 
elements of human thought, even the basic ones, until we can do 
something a lot simpler and represent the reality in which a business 
exists. And similar challenges for the reality in which buildings exist 
and so on.    (05)

This is why I do not think that a foundation vocubalary for a 
human-language dictionary like Longman is a useful reference point for 
helping machines that belong to businesses to talk to one another, 
though I can see that a lot of very usful ideas come out of it.    (06)

A few people have commented along the way, that these top level 
categories are not useful. However that doesn't give me much to agree or 
disagree with, so perhaps I can outline some brief thoughts:    (07)

Independent/Relative/Mediating:    (08)

These categories have been figured out by a whole range of philosophers 
under different words. John Sowa gives a better explanation than I could 
both of the history and the thinking. My practical engineering take on 
this is that here is a set of partititions which is well attested, which 
works, and which can be explained to anyone very simply (by principle or 
example or both).    (09)

It also directly solves real business problems, though a good database 
engineer should not need to call in an ontologist to figure out that 
clients, counterparties, underwriters of a security issue and so on 
should not each be defined with a phone number and an address. However 
I've seen database models in the financial sector which attempt to do 
just that. So I have used these concepts to explain to them how a better 
conceptual model would avoid such things.    (010)

So this is just plain simple engineering: a Relative Thing is a concept 
which is only meaningful in relation to some context (the Mediating 
Thing, for example Securities Issuance). Each Relative Thing also has a 
single, simple relationship to some kind of Independent Thing - the 
business, person or whatever that actually has an address, a bank 
account and so on. By the way I'm using capitals to denote classes of 
Thing.    (011)

The other two I agree are less clear cut, and in both cases they also 
represent some specific ontological commitment which would rightly vary 
from one ontology to another. However, having a partitition for 
Occurrent Things is very useful for representing things which are an 
Event, an Activity, a Process and so on. You would not get far modeling 
things about financial securities for example without some way of 
modeling those. However, to align such a model with other ontologies one 
would have to find some way of relating the different kinds of time 
scales against which something is defined as an occurrent or a 
continuant thing (for instance the reign of a king can be described as 
both). Perhaps relating them all to some 4D framework is the answer 
here, and if such relation can be done without breaking the individual 
ontologies with their different timescale-based ontological commitments, 
that would be good.    (012)

Similarly there are different ontological commitments likely for 
different applications in the distinction between concrete and abstract. 
Most people would regard only things you can weigh and kick as concrete, 
but for example I have chosen to define financial securities as concrete 
things whether or not they are in material form - a partition that 
artificially separates similar financial securities based on whether or 
not they are made of paper would not be useful IMHO.    (013)

However certain abstract things are clearly abstact no matter how you 
look at them. In business terms, things like goals and strategies would 
be abstract but necessary (an Actor pursues some Goal in carrying out 
some Activity). Here I think the approach you have developed in COSMO 
could make a good foundation for certain classes of Abstract Thing that 
are also relevant to businesses. Perhaps the labels you use at the top 
of that particular part of the model would be more useful and meaningful 
than saying "Abstract".    (014)

> Roles, for example, can cut
> across several disjoint categories.    (015)

I don't get that. I understand that in setting out to create an ontology 
from scratch, one can more or less decide whether it's more appropriate 
to define a role as a thing or a relationship. To me it makes sense to 
have a second order thing (Actor and/or Party - I use both as they have 
distinct and useful meanings), and to have a "role" relationship to that 
second order thing. You end up with more different terms than perhaps 
one database or applicaiton would need, but you can (hopefully) relate 
it to a full range of data models without it falling over. See for 
example securities issuers, Special Purpose Vehicles (SPVs) that are the 
issuers of certain types of collateralized debt obligation and so on - a 
difficult mess to unpick but I think that separating roles and relative 
things gives a rich enough vocabulary to start to unpick it.    (016)

This is also where I find it helpful to define a "most general" set of 
classes and relations, such that every Activity has an Actor who has a 
Role, with all roles being specialised meanings derived from that most 
general archetype of a role. So the role of "manages project" is a 
specialised variant of the concept of Role, as applied to the Actor 
which is Project Manager.    (017)

In terms of that I really don't understand what you mean about roles 
cutting across several disjoint categories.    (018)

> If you have some specific logical properties that you think should be
> assigned to those top-level categories, send me a note and I will be happy
> to see how they can be related to the COSMO elements.
The thing about high level categories is that I would not expect to see 
them have many logical properties, otherwise I would suspect that there 
was a thing above them in the taxonomy that did not have those properties.    (019)

Relative Thing has an identity which is an Independent Thing, and a 
context which is the Mediating Thing.    (020)

Occurrent Thing - I've defined Event as a sub-class of this, and it is 
something that has a time and a place (incidentally some ontologies e.g 
INDECS call this a "context". I was reluctant to assign these specific 
properties to Occurrent Thing itself, so it has no properties, but I'm 
open to wiser consel on this one. After all, a process is also a kind of 
Occurrent Thing, but it is not a kind of Event, I think. Likewise a 
State is certainly not a kind of Event but it belongs in the Occurrent 
partition.    (021)

> There is a lot of work already done in the OpenCyc, SUMO, DOLCE, BFO,
> ISO15926, and other ontologies that can serve as a starting point if one
> wants to redo the top-level categories.  Seeing how these can be related to
> each other is quite educational.
Indeed there is. My first impression was that the Cyc / SUMO etc. views 
went too quickly into domain specific partitions. I realise that the 
reasons have been well explained here and I need to spend more time 
looking at this, but I still think that certain concepts - specifically 
legal, financial, business (transactions etc. as per REA) and so on are 
relevant to every business (but perhaps not to every building, every 
organic molecule etc.) and there is I believe a real case for some 
common terms in the business world. The terms one would need should 
mostly be available in standards by now, if not in ontologies.    (022)

However I am also hesitant to use any ontology where the classes cannot 
be traced directly back to some real industry standard, since although 
those may not (yet) be ontologies for the most part, they should have 
figured out all the meaningful concepts that are useful in their various 
domains of discourse.    (023)

Best regards,    (024)

> Pat
> Patrick Cassidy
> MICRA, Inc.
> 908-561-3416
> cell: 908-565-4053
> cassidy@xxxxxxxxx
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/ 
> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>       (025)

Mike Bennett
Hypercube Ltd. 
89 Worship Street
London EC2A 2BF
Tel: +44 (0) 20 7917 9522
Mob: +44 (0) 7721 420 730
Registered in England and Wales No. 2461068    (026)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (027)

<Prev in Thread] Current Thread [Next in Thread>