Re: [EXT] Upper ontologies

Hi Anthony,

One can refine the 'fundamental' relations (though there is no agreement on what they are) and add sub-relations as finer-grained ones cf all on par or more/less compact as variations in encodings, but even if you were to agree on that, then that still does not resolve  disagreements as to the nature of what a relation is and those distinctions still would result in different upper ontologies.  The two well-known ones in use are called 'standard view' and 'positionalist'.

The standard view is baked into in OWL and first order predicate logic: there's a relation/relationship/objectproperty and a fixed order to its participants. in shorthand notation, e.g., for loves(John, Mary) or, at the type level, partOf(Heart,Human), then John has to take up the first place in the relation and Mary the second, not the other way around--that John loves Mary doesn't mean that it is reciprocated in that loves(Mary, John) would hold as well, and likewise for partOf. When the participants are swapped in their order, we seemingly have to have another relation to stay true to the state of affairs, say, lovedBy(Mary,John) and hasPart(Human,Heart), respectively.

It's easy to declare loves & lovedBy and partOf & hasPart inverses of one another, sure, but one can argue (among other reasons) that it's not right to need two relations for the same state of affairs between John and Mary or between Human and Heart. Instead, there are two ways to talk of the same relation that holds between its participants. One of the ways to address it is positionalism.
With the positionalism, there's only one relation--say, loving and parthood, respectively--and it relegates loves, loved by, part of, has part and so on to a language layer. That relation in positionalism consists of n unordered roles in the n-ary relation, where each participant plays a designated role. Indicating such roles within square brackets, then writing it down as
loving([lover],[beloved])
or
loving([beloved],[lover])
is all the same, but we need a function that assigns John to [lover] and Mary to [beloved]; e.g., play(John,[lover]) and play(Mary,[beloved]). One can do likewise for parthood and any n-ary relation (alike R([role1],[roleA],[role-i])). Alternatively, it can be seen as taking projections over the participants in the relations. It's a common notion in UML class diagrams with its association ends and EER with the non-directional relationships and its relationship components, as well as in ORM (Object-Role Modelling).

Positionalism asserts that there are such roles that make up relations and that they are essential components of relations, whereas the standard view does not admit them into the universe, which is irreconcilable ontologically. This is an orthogonal issue to the 3D/4D distinction that Pat described, so there we go with adding more to the proliferation of upper ontologies.

Regards,
Maria


----
Dr. Maria Keet
Associate Professor
Department of Computer Science
University of Cape Town
Cape Town, South Africa
tel: +27 21 650 2667
fax: +27 21 650 3551
email: mkeet@cs.uct.ac.za<mailto:mkeet@cs.uct.ac.za>
work: http://www.cs.uct.ac.za
home: http://www.meteck.org


On 20/01/2021 17:40, Anthony Moretti wrote:
CAUTION: This email originated outside the UCT network. Do not click any links or open attachments unless you know and trust the source.

Thanks for the example Marcel. Can’t the relations from the simple representation and the complex representation coexist under the same hypothetical “SUO” though?

Anthony

On Wed, Jan 20, 2021 at 8:46 AM Margaret Warren <mm@zeroexp.com<mailto:mm@zeroexp.com>> wrote:
That's a great idea, Dan..I would be willing to work on this.

-------- Original message --------
From: Dan Brickley <danbri@danbri.org<mailto:danbri@danbri.org>>
Date: 1/20/21 09:31 (GMT-05:00)
To: Jos De Roo <josderoo@gmail.com<mailto:josderoo@gmail.com>>
Cc: Hugh Glaser <hugh@glasers.org<mailto:hugh@glasers.org>>, Patrick J Hayes <phayes@ihmc.us<mailto:phayes@ihmc.us>>, Semantic Web <semantic-web@w3.org<mailto:semantic-web@w3.org>>
Subject: Re: [EXT] Upper ontologies

If it were in a wiki somewhere it could approximate a book...

On Wed, 20 Jan 2021 at 13:45, Jos De Roo <josderoo@gmail.com<mailto:josderoo@gmail.com>> wrote:
Did exactly the same, appended Pat's posting to local file pat.txt
Thanks Pat !!!

Jos

-- https://josd.github.io/<http://josd.github.io>


On Wed, Jan 20, 2021 at 1:34 PM Hugh Glaser <hugh@glasers.org<mailto:hugh@glasers.org>> wrote:
Thanks Pat - an excellent and well-timed posting.
I will save it for future use and savouring.

I think that the discussion illustrates the problem.
Upper Ontology is a concept.
Some people conceive of it as singular.
Others as plurality.

> On 20 Jan 2021, at 06:17, phayes@ihmc.us<mailto:phayes@ihmc.us> wrote:
>
> OK, I had promised myself to stay out of these discussions, but …
>
> No, this will not work. It has been tried, many times. Every existing upper ontology was built by people who honestly believed that they would do this, and were willing in some cases to sacrifice years of their professional lives to achieve this. I was part of several of these initiatives, some of them financed by agencies like the US Army and DARPA. But still we have a host of upper ontologies.
>
> And there is a good reason why this happens. Yes, we are all talking about the same one world. And let us assume, for the purposes of argument, that we are all using the same formalism. (Of course not true, but translating between formalisms is relatively straighforward.) Still, we will not all create the same ontology, or even compatible ontologies. (I called this the "diamond of confusion" in a talk about 20 years ago.) And this is because an ontology is, in Tom Gruber's phrase, a formalization of a /conceptualization/, not a formalization of /reality/. And while there is widespread agreement on the nature of the actual world, there is most emphatically not universal agreement on conceptualizations of it. People are still arguing about ontological conceptualizations that were discussed by the Greek philosphers 2000 years ago.
>
> I can illustrate this with a very old, /very/ thoroughly discussed example, which is how to describe things that are extended in time. That is, things in the physical world, not abstract things like numbers or ideas. There are two main ways to think about this.
>
> In one, often called the 4d perspective, all things in time and space occupy some chunk of time and of space, and we describe them by talking about their parts, including their temporal 'slices'. So I – PatHayes4 – am a four-dimensional entity, and we can say things like [**]
> Weight(PatHayes4@2020) > Weight(PatHayes4@1966)
> to express the regrettable fact that I am getting heavier. The @ symbol here is a function that takes a time-extended thing (me, in this case) and a time, and returns a time-slice of that temporally exended thing. So PatHayes4@1966 is a thing that I might call 'Me in 1966', and PatHayes4 is me throughout my lifetime. The me who is present at any particular time, such as now, is only one momentary timeslice of the entire PatHayes4.
>
> In another way of thinking, there is a fundamental distinction between 'things' (like you and me) and 'events' which happen. (Other terminologies are often used: continuants vs occurrents or perdurant vs endurant. I will stick to things and events.) Things are 3-d, dont have temporal 'parts', and are identically the same thing as time passes. (They continue as time passes; they endure.) Events happen, are temporally extended and have temporal parts. In a nutshell, things are 3-d, events are 4-d. So a football match, a wedding ceremony, a theatre performance are all events, but the players, guests and actors (and many other things) are things. And a guest at the wedding just as he arrives is identically the very same thing as when he is going home after the wedding, though his properties may have changed. Time parameters are typically arguments of properties rather than attached to names, so that my getting fatter might be written
> Weight(PatHayes3, 2020) > Weight(PatHayes3, 1966).  Note that the first arguments of these two are identical.
>
> I will not go into the pros and cons of these perspectives. Each of them has been a foundational perspective for an upper ontology in widespread use, and each has been successful. Users and proponents of each have published detailed philosophical defenses of them and critiques, sometimes bordering on slander, of the other. Each of them "works". But they are profoundly incompatible.
>
> The problem is that the 'things' of the second perspective are /logically impossible/ in the first perspective, since they have no temporal parts or extents – they are purely 3-d. So the thing PatHayes3 cannot be identified with PatHayes4. But it also cannot be identified with any particular 'slice' of PatHayes4, since these have different properties, but PatHayes3 is identically the same thing at different times. There simply isn't room in the 4d ontology for things like PatHayes3 which have no temporal extent yet exist at different times. So, one might respond, the worse for 3-d things: but in the second perspective, those 3-d things are the basic fabric of reality, so wthout them there cannot be any events to happen to them.
>
> This incompatibility is not just a philosophical issue: it has ramifications all through the ontologies, affecting how entities must be classified, the syntactic form of the sentences that describe them, even how many of them there are. People learning how to use these ontological frameworks have to learn to /think/ in distinctly different ways.
>
> As my friends know, I could expand on this topic at much greater length, but maybe this will serve to give an idea why the naive idea of just 'choosing the best pieces' of a variety of upper (or lower, for that matter) ontologies is not going to work, any more than trying to make a hybrid car by just taking the best parts of Ford Tbird and an electric golf cart.
>
> There is a reason this field is called 'ontological engineering'.
>
> Pat Hayes
>
> [**] This fragment of formalization is absurdly simplified, but it captures the heart of the matter.
>
>> On Jan 18, 2021, at 8:49 AM, Mikael Pesonen <mikael.pesonen@lingsoft.fi<mailto:mikael.pesonen@lingsoft.fi>> wrote:
>>
>>
>> This is the way I see it too, if there would be effort for the common UO. Take the best parts of the existing UOs and harmonize them.
>>
>> One would think it would also save some work in future for anyone making domain ontologies. Just choose the best point of view from “Standard Upper Ontology” and start building on it (if there were more than one point of view available in "SUO").
>>
>> On 17/01/2021 3.46, John wrote:
>>> I think the issue of upper ontologies could be relatively straightforward. Some esteemed organization (W3C?) should initiate an upper ontology working group that would become a major effort. By major effort I don’t mean going to the moon or Mars, but something very major indeed. It would probably require funding from multiple governments to reach the necessary scale of effort. It would select an eminent group of experts as the core working group members who would have the final say in defining the “standard upper ontology”. Inputs would be requested from a very wide source of developers to be considered by the working group. Th e goal of the working group would be to identify, as best as possible, what is true and meaningful in terms of relationships and what is not. A good starting point would be measurements and geographic classes and properties. There is a lot of good work already in these areas that could be leveraged. The next job would be to identify a constrained list of the top-level real world things that most domain specific ontology would need to reference. The ultimate release of the “Standard Upper Ontology” would serve the widest categories of ontology developers and they would all be strongly encouraged to use the standard in order to achieve the maximum interoperability. Those ontology developers who simply cannot live with the standard could go there own way, but realizing they have given up the opportunity to seamlessly interoperate with the majority of the Semantic Web community.
>>>
>>> John Flynn
>>> Semanticsimulations.com
>

--
Hugh
023 8061 5652



Disclaimer - University of Cape Town This email is subject to UCT policies and email disclaimer published on our website at http://www.uct.ac.za/main/email-disclaimer or obtainable from +27 21 650 9111. If this email is not related to the business of UCT, it is sent by the sender in an individual capacity. Please report security incidents or abuse via https://csirt.uct.ac.za/page/report-an-incident.php.