Re: [ontolog-forum] language vs logic - ambiguity and startingwithdefini

[FERENC KOVACS <f.kovacs@xxxxxxxxxxxxxx>]

Doug,

(My apologies to Peter. Perhaps we could switch to private correspondence not to annoy the rest of us.) as now I have your email adddress

 

As I expected we have difficulties arising from using a different paradigm to se the world. Again, I remove anything but what to respond as an outstanding issue. Thanks a lot for trying to make sense of what I write

I totally misunderstood you.  I now take you to mean that "the three
core ontology concepts are the concept of object, the concept of
property, and the concept of relation."  Is this correct?

Yes, it is. But I also claim that the referents of these concept exist, as objects http://en.wikipedia.org/wiki/Object_%28philosophy%29 wich originally meant in Latin something blocking your vision to see the whole picture, an issue if you beileive I reality or not. This object is then processed by the senses and by mental operations to derive the rest of the CHAIN of associations in an effort to understand the world and discover RELATIONS.

Reading on, i'm starting to wonder if you are defining what would be
the core concepts for a natural language processing (NLP) ontology.
Many of the statements below seem to me to be based on the assumption
that the topic is NLP.

Yes, you are right again. But In my view there is no separate ontology fo NLP and atronomy and biology for instance – as regards its top level and the mental operations that are in  place to produce them. We allé have the saem faculties, but different knowledge due to different experience in life.

> > Is the distinction between properties  and relations that
> > properties relate objects to datatypes while relations
> > relate multiple objects or have more than two arguments?
>
> We are nowhere near dabases yet.

I was not referring to databases at all.  Although databases columns
restrictions can be defined in terms of ontologies and rows can be
mapped into one or more assertions using ontologies, discussions of
ontologies should not be interpreted as discussions of databases, imho.

 

But you have introduced dabases in your passages whereas I am trying to start discussing a subject by defining my terminology step by step so that we can verify each definition or sense.

> The distinction between properties and relations
> is that properties are created in a relation
> (which contains a verb element) of an object with another object.

I find this confusing.
* Properties are created in a relation between two different objects.
* A relation contains a verb element.

 

Looking at the world what you get is the result of chunking. The biggest chunk (with a boundary of varying size) is the universe (quantity one) and its properties existence (quantiy one). We can separate ourselves from the environment, so a small chunk is a man, and a man is further divided as a unity of body and soul (in a hazy relation mind and spirit)  Thus it is a man (object no 2, therefore subject, becasuie it is smaller than the other, it is subdued) mentally processing (relating, IS in a relation to – this is a verb no matter which version you like) the universe (an object no 1) as a result of which object no 2 infers that both objects (no1 and no 2) exist i.e. they have the property of existence. This reality check is always necessary.

Verbs seem to me to be language features, not features of an ontology
or of the world to be described by an ontology (unless the topic of
the ontology is language). 

 

Well, that is the problem with your(not personally yours!)  current ontologies.

 

Verbs often are used to represent either static situations or events (which can be considered to be active situations).  Situations, thus, would be parts of the universe to be
described, requiring "Situation" to be a key ontological concept.

 

Neither situation nor event are suitable for identification of the parts of the world, in other words chunking a universe. I would get rid of them once for all.

But back to your description.  Are you defining a "relation" to be
a predicate (and thus a mental construct) and a "property" to be
an assertion connecting the predicate to multiple objects?

 

Why do you think it is only relation that is a mental construct, and property, assertion and multiple objects are not?

> Events and situation are not core concepts, they are derivations
> (and may not be very useful in this model).

OK, so you define them as derivations, i.e. subtypes, of object?

 

Not quite. I have not been using type, token, etc. nor subtype.

> The concept of object, just as object in physics is
> seen in a couple of templates. Form and content is just one of them,
> another semantic primitive

"Form and content" is a single semantic primitive?

 

Object is a semantic primitive (yet a product) and an object is seen as a unity of form and content (two facets if you like) – properties showing in two different relations (mental operations, verbs) yielding two different qualities (namely form and content)

It seems to me in your ontology of objects, relations, and properties/
statements, "form" would be one relation and "content" would be another.

 

Nearly, see above. How did statements get in there?

So what you are proposing is an ontology of NLs, not an ontology of
the world known to us.

Why do you think that these two are different pots of jam?How do you share your knowledge of the world with me?

If we are to understand a statement about the world by mapping it
through a process to model an NL, it would greatly complicate the
process, imho. 

 

It is just the other way round. It is an NL statement you have to analyse semantically and describe in a similar fashion to BNF in a process of recursion. Obviously the same semantic content may be represenetd not just by different words, but different word clusters and /or statements

 

That seems to me like understanding Finnish by
mapping it into English and then trying to understand the generated
English instead of understanding the Finnish directly.

 

Then I am not clear enough. The objects in the world (in spacetime) are given names in various languages and countries. It is the chunking that matters to maintain the identity of the sbject matter that are spoken of in different lingos. We tend to forget that thinking is not limited to high-tech societies, animaly and even babies can think. For some reason we assume that thinking is done in terms of concepts which by necessities have to be named and sometimes formalized.

 

I would suggest trying to view the ontology as another language & having
reasoning being done in that language instead of having to convert
it to and from an NL.  Certainly input & output for a semantic system
may use a different language (a DB language, a CS language, or maybe
even an NL), but that input would be converted to the semantic
language and ontology and processing would be done in that form only
converting back to a different language for output.

The semantic language you are talking about is only good for computers Most of us cannot make sense of it as we have different processors in the mind.Why do not you want to know how your own processor is contructed – by looking at its set of instructions?

Yes, I am talking about the creation process, in which we produce
> concepts and data among other things,

These are two very different things.  One can view creating the concepts
as authoring a program and creating the data as using the program.

 

I am talking about the creation process without thinking about a computer. In this simple model a creator (object) creates (relation) a creation (property). But creation as property is possible to see as an object, just as creation as a verb is possible to see as an object too. Before you find that mind buggling I need to remind you that our mind does something we (Hungarians) describe intuitively as rotation of objects in space. Obviously if you de that exercise you find nothing surprising in looking at thingies from dfferent angles.

> which are products also in the sense that they need to
> be complete when finished. Complete means a whole, something like an
> integer.

A very limited definition.

As it is in here: Object=::object/relation/property

With this argument, the verb can be anything.  Amato ergo sum.

Any verb you mean.

And i must conclude that the property of something i see on a movie
screen is existence?

 

Among other things subject to your perception and state of mind. You may also believe (as the first time movie goer) that there are people on the screen  whose heads are cut off and you run out from the show.

You are obviously discussing NLP, but framing it in terms of a
discussion of ontology.  I think it would be easier for all concerned
to understand the discussion if the distinction were made and
it be clear to the reader which of the two fields is being discussed.

Well, hiow do you want to have a proper SUMO and a normalized semantic enterprise architecture if you keep your knowledge of the world and that of NLs apart?

> If so, i would suggest that this "explosion" could be carried  out
> byformal
> logical operations -- they wouldn't need to be mental  operations.

Certainly generating statements that can be derived from a set of
statements expressed in a logical language could provide such an
explosion.

Unless you have an algorithm with an end state -  as you would in reality. If you like metaphores and if you are familiar with the Bible, you would agree that in the process of genesis the end of process was that God saw that it was good. Go to start of cycle.

 

I interpreted, "verbs are seen as the representations of relations",
that way.

You tell me what else verbs represent, please.

> What verbs do in English for example is to make a message out
> of a cluster of words. To make a message you need a clause.

I would suggest that verbs make statements, not messages.  I consider
a message to need a sender and a recipient.

 

A satement needs to be interpreted semantically. This is how you arrive at messages that make sense and messages that do not. Of course you have a sender and a recipient whether explicit or implicit.

> Only clauses make sense in NLs.
> (Further reading : David Crystal: Making Sense of
> Grammar, Pearson Education, 2004, UK ISBN 978-0-582-84863-4) No further
> grammar/syntax analysis please for the time being.
>
> > Features of the events/situations can certainly be modeled by relations
> > that ignore the events themselves, but this would require multiple
> > relations to be defined for each verb depending upon what other phrases
> > happen to be in the sentence.  Using this technique makes it difficult
> > to add more information about the same event and requires multiple
> > rules to inter-relate the multiple relations that represent the same
> > verb.

> In contrast to noun phrases that basically refer to spatially related
> objects
> whether real or virtual, and are titles, headers, labels or tags

Many objects (bank accounts, names, words, songs, laws, game rules, ...)
are not spatially-related objects.

 

They are, because the huerarchical structure they are arranged in IS a spatial representation. Most of these terms show contained in, or Boolean relations that are abo ovo derived/abtracted from the observation of objects in space


Since no context is given the described event is not unambiguous.  But
most uses of NL are ambiguous.  The ontological representations would
be unambiguous.  Given sufficient context, the statements above could
be mapped into unambiguous, but underdefined, descriptions in a knowledge
base or be generated from a representation of an event in a KB.

 

So you agree that meaning cannot be defined without context and they are inversely related.

A throwing event is an "action" in which a mobile physical object propels
a smaller and (normally less massive) physical object through space either
with a hand at the end of a arm or a physical structure approximating it.

 

Now is this defintion precise enough to tell me how to find it semantic terms?(object, property and relations for instance?)  Can you identify it in terms of spacetime as you can do in case of non moving objects? Can you be specific enough to turnyour definition t into a piece of data to act on?

What is in someone's memory or on film may be a partial representation
of the event.

 

You may call that that way. But in my interpretation “event and partial” are not very useful.

> because for you the only semantic property is truth.

Please do not attribute to me properties you only guess.

You are totally wrong with this guess.

I am sorry, I did not mean you, but the people in the forum wjho believe that semantic anaysis is about using formal logic.

> My point is that ontologies need to be integrated and not merged.

? You were discussing dictionary definitions (which can be source
material when creating an ontology), NOT ontologies. 

 

Yes, but “by the way” ontologies have similar problems as far as I can tell from the posts here. The name SUMJO suggest that you are after merging items that are not compatiblenot just because of the difference in domain, but the in the logic used to produce them. Logis in the sense of mentsl operations used that the experts are not aware of, especially with respect to emergence or origin of terms as no algorithm or temporal sequence is represented.

Legacy data bases often allow new inputs in fields (e.g., occupation)
that were previously restricted.  If such a DB were mapped to a KB,
an associated ontology would have to be expanded to include the expanded
range of the field.

Good for them.

> > Here, i assume you mean grammatical person, not requiring the message
> > to relate to a person.  Person, in this sense, is a linguistic feature
> > of sentences of many languages, not necessarily relating to a feature
> > of the meaning being discussed.
>
> Yes, you are right on the first sentence. A message relating to a grammar
> person does also relate to a person per se.

?  These two statements contradict each other!

Counterexamples of the second sentence:
* The cat jumped on the bed.
* The hurricane broke the window.
* The book fell off the shelf.
None of these relate to a person per se.

A slight misunderstanding here. A grammar person (subject) equals an object, a person per se that may not be a good way to put it.

 
Regards,
Ferenc

_________________________________________________________________
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    (01)