RE: Question about ontology

Hi J,



My reaction inline.



Cheers,



Paul



*From:* jmcclure@hypergrove.com [mailto:jmcclure@hypergrove.com]
*Sent:* dinsdag 10 september 2013 19:38
*To:* public-owl-dev@w3.org
*Subject:* RE: Question about ontology



Well sure a statement can be actively or passively stated for instance. And
there's no dispute that the physical organization of a set of triples can
be altered without changing the semantics of its statements (but so what?).

That wasn’t the point I was trying to make.

I do strenuously dispute though that you can "turn adjectives nouns and
verbs into each other".

Would you say then that the following pairs of sentences essentially mean
different things?

·         “Johan and Mary are married.” versus “There is a marriage of John
and Mary.”

·         “John drives the car.” versus “The driver of the car is John.”

·         “That is a 30 feet high house.” versus “The height of that house
is 30 meters.”

·         “The liquid is in a bottle.” versus “The liquid is bottled.”

My point remains that the lexical category of a word is not essential for
the meaning (concept) behind the word.

And I strenuously argue that any ontology that ignores adjectives and
prepositions is seriously, deeply flawed -- such an ontology makes
assertions such as yours ("*no difference*...(between)... a
relation/predicate and (its) object") almost reasonable!

I am not saying that ontologies should ignore anything. I am just saying
that one should look through (or behind) the lexical category of a word.
Ontologies shouldn’t model adjectives, nouns, verbs or any other lexical
category, they should model concepts/units of meaning to begin with. The
words are in there “just” to get the concepts across to humans. Words are a
**vehicle** for meaning, not a **container** of meaning. The lexical
category of a word it gets from the more or less arbitrary grammatical
context of the rest of the sentence.

My problem is that there IS a difference. Objects are nouns. Subjects are
nouns. But predicates can NOT be nouns; nouns can NOT be predicates.
Rather, common nouns connote types of things: they are but labels for
distinguishing generic identities. Nouns are not candidates for predicates!

The only reason why predicates cannot be nouns is that predicates are
semantic and nouns are lexical. But, nouns can most definitely be used for
indicating predicates, as verbs can. And, preciates can be reified to
become objects. From what you are saying, I understand that you use an
intensional paradigm for looking at the world, like the object-oriented
paradigm does. There is nothing wrong with that, but my point is that
that’s a **choice**, it’s not “given by nature”. It is very well possible
(as well as practical) to do without that choice.

But what do we see as a consequence of using nouns for predicates? Well,
first off, you see tendencies to define property hierarchies that DUPLICATE
noun hierarchites -- same noun, first upper cased in the noun hierarchy and
then lower-cased in the property hierarchy -- golly how pregnant with
meaning that first letter is made to be!

I don’t see why this would be a consequence. I just take the freedom to use
that lexical category for the name of a concept that communicates best in a
given situation. I need no duplication. Why would I? If I have modeled
illness, I have modeled ill. It’s the same concept for a start. If a person
has it, I can call him ill. But if a doctor studies it, I’d probably call
it an illness.

And even if you think that “ill” and “illness” are essentially different,
you must agree that they at least share meaning. I’d rather prevent myself
from having to introduce zillions of rules like “IF X has an illness THEN X
is ill.”

How can the outcome of a massively bloated noun-infused property hierarchy
be said correct? Ever heard of Occams Razor?

Good that you mention Occam’s razor. It supports my point. In intensional
paradigms, there are upfront assumptions, such as:

·         objects cannot be properties

·         objects cannot be processes

·         properties cannot be processes

My point is: you can perfectly do without these assumptions and still
produce precise (formal) models. In fact, it is possible to have a
single-construct language (that’s about as concise as your meta-model can
get) that combines and significantly extends the expressive power  of OWL,
SBVR, object-orientation, workflow models, whatever. The trick is to (1)
postpone such assumptions and (2) introduce
contextuality/intersubjectivity. This implies a shift to the intentional
paradigm (with a t, that is, not an s), which is wider that the intensional
(as well as the extensional) paradigm.

In doing so, you don’t lose ground, nor precision.  Rather reversely, you
gain scalability, managed relativity, and elegance. And, because you can
reintroduce these assumptions into your model whenever you like, you don’t
lose expressive power. What you have to give up though, is the **upfront**
requirement that there can be only one-truth-per-model, and the **upfront**
requirement of automatic reasoning. Notice the “upfront”. If you like you
can restrict your model so that these properties reappear.

(Yes, we have applied it in practice, for clients. And no, it is not yet en
vogue.)

Cheers!

Paul

thanks /jmc



On 09.09.2013 23:06, Paul Oude Luttighuis wrote:

Hi J,

 I’m afraid that’s not so fundamental. We shouldn’t over-interpret natural
language grammar (nor physical appearance for that matter) when it comes to
semantics. I can easily juggle around with adjectives, nouns and verbs and
turn them into each other without changing the meaning of the sentence.
Natural language grammar is superficial/superimposed structure for a large
part.

As much as that I can say that I am **in** a marriage (which I am), I can
say that some marriage relates me to my wife. Also, there is no fundamental
difference between a relation/predicate on the one hand and an object on
the other. They are just different perspectives of the same. A marriage is
equally a relation as it is an object. We tend to take them apart, but for
superficial reasons (natural language grammar, or mathematics). Reification
is no exception, it’s the rule. Don’t we say that we take **part** in a
relation?

So, I see no problem in having “full bottle” as the relation in this
example. It relates “empty bottle” and “liquid”. The pouring is the context
responsible for defining that relation. In case we can’t express
contextuality, I would settle for seeing it as the third component of the
relation. This makes sense (in this example), because you need “pouring” **
first** before you can have full bottles.

 By the way, by saying that the difference between **being in something**
and **taking apart in something** is superficial, I do not say that such
distinction is always irrelevant. But, in case I would feel that this
distinction is relevant, I should think about what makes the difference.
This context (the differentiator), though, is **specific**, it’s not
fundamental. This even holds for the physical context.



Cheers!

Paul





*From:* jmcclure@hypergrove.com [mailto:jmcclure@hypergrove.com]
*Sent:* maandag 9 september 2013 22:09
*To:* public-owl-dev@w3.org
*Subject:* RE: Question about ontology



Well... If 'part-of' applies to liquid in a fullbottle, then wouldn't it be
true you'd say that some car is part-of a full-garage? No, some liquid is
IN a bottle, and some car is IN a garage. There's no need to redefine our
most fundamental speech patterns!

/jmc

On 09.09.2013 11:52, Paul Oude Luttighuis wrote:

Dear Sybri team,



My first question would be the question of existence-dependency, in other
words: which terms are needed to define others? In this case it seems to me
that “empty bottle”, “liquid”, and “pouring” all preceed “full bottle”.



The semantic structure at hand then is: the context of “pouring” defines
“full bottle” in terms of “empty bottle” and “liquid”.

In other words: pouring **is** not a relation between empty bottle and
liquid, it **defines** such relations.

Such contextuality however cannot be expressed by OWL-type ontologies.



If you would however want to stick to the OWL world, then part-of would
work perfectly I guess, but then I would see “pouring” as part-of “full
bottle” as well. There is no upfront semantic reason not to.



Regards,



Paul



*From:* System Bridge [mailto:sysbri1@gmail.com]
*Sent:* maandag 9 september 2013 13:32
*To:* public-owl-dev@w3.org
*Subject:* Question about ontology



Hello,

we are group of PHD students and we would like to ask you (ontology
experts) for help/advice.

We`re trying to make a simple expert system using ontology as knowledge
base. We have come to few problems and before making any conclusions, we
would like to confront it with you.



We noticed that every explanation and example we found uses object
hierarchy, e.g. OneThing isPartOf OtherThing or OneThing hasPart
OtherThings. We don`t know how to model process which also causes that
resulting object will be assembled from some other objects. For example:

   a) Object Empty bottle

   b) Object Liquid
   c) Process: Liquid will be poured into the Empty bottle and thus will
create some new object Bottle filled with liquid - see image attached.


What we need is to define a relation “Pouring” that is between liquid and
empty bottle. In fact we don’t really need “is part of” relations if there
is a way to express “is part of” implicitly in “Pouring” relation, because
it is obvious that the "Bottle with liquid" was created by "Pouring" the
"Liquid" into the "Empty bottle". Also the direction of "Pouring" is
important for us.

So, the question is whether you may help us either by explaining this
particular example or providing us with helpful source of information how
to solve it.

Thanks in advance

Your sincerely

Sybri team, University of Zilina, Slovakia[image: Inline image 1]