Re: THEORY nouns and cases (was: Verbs derived from noun cases)

Philippe Caquant said:
>
> My idea to figure logically an idea like "Rex is a
> dog" would be something like:
>
> Rex ---- RELATION ---- dog
>
> (which is exactly the case in English or in French,
> only the relation expressed is a very vague one: "to
> be", which can mean a hundred different things; while
> in Russian, the copula wouldn't be expressed at all,
> or only typographically: Rex - sobaka).
>
> So the point would be: what kind of conceptual
> relation is this ?
>
> And my answer is: the relation is "is-an-instance-of",
> which is exactly the same as when you say "Tony Blair
> is a man", or "Saturn is a planet"; but NOT, for ex,
> "A rose is a flower", and yet much less, "Rex is
> brown", "Rex is dead", "Rex is Lassie's son", or
> whatever.
Ah. My mistake, I somehow thought we were talking about natlangs.

The Waldzell ontology should work well for what you're talking about.

In short:

Every open-class name refers to either an instance or a class; every
instance is either an object or an event; and every class is either an
object class or an event class.

One class can be a SUBCLASS-OF another class; it follows that every class
ultimately derives from either the class of all objects or the class of
all events.

An object or event can be an INSTANCE-OF a class.

An object or event can be a PART-OF another object or event.

An object can PARTICIPATE-IN an event.

One event instance can INFLUENCE another event instance.

One instance or class can be IDENTICAL to another instance or class.

Classes can be defined using any propositional (AND, OR, NOT) combination
of these relations.


"The cat sat on the mat."
This sentence refers to an event instance E for which the following
assertions hold: An object instance C, which is an instance of the object
class CAT and of the class of all objects identifiable by both speaker and
hearer in the situation of this utterance, participated in an event
instance S (a) which is an instance of the event class SIT, (b) which is
an instance of the class of all events that began prior to the time of
this utterance and (c) in which an object instance M, which is an instance
of the object class MAT and of the class of all objects identifiable by
both speaker and hearer in the situation of this utterance, participated.

E IDENT {
  C instance-of CAT and
  C instance-of IDENTIFIABLE and
  C participate-in S and
  S instance of SIT and
  S instance of PAST-START and
  M participate-in S and
  M instance-of MAT and
  M instance-of IDENTIFIABLE and
}

(This is an on-the-fly analysis and is not intended to make any claims
about specific details of English semantics; it's just to illustrate how
the ontology works.)


I'm working on a book-length treatment of the ontology of semiotics that
uses the above as its upper ontology; details of the latter have evolved
somewhat since its only other (rather ancient) prior publication to the
Web at

     http://www36.pair.com/waldzell/Canon/ontology.html


> If you want to build an automatic translator between,
> say, English and Tagalog, or Hungarian and Cherokee,
> so you have to proceed more or less like this:
> 1/ Convert the source speech into an inner conceptual
> representation, using a particular set of rules
> 2/ Convert the inner conceptual representation into
> the target language, using another particular set of
> rules.
I disagree. Whether for translation or any other purpose, source speech is
not "converted" into anything. It has to be _understood_ to be useful.
That's why machine translation is still so difficult: computers can't be
made to understand very much (although we can make them do conversions out
the wazoo). When all you have is a hammer, everything looks like a nail --
so the AI folks try to make do with conversions even though understanding
is the only thing that will really fit the bill. That's one of the reasons
I got out of the AI business and into the Cognitive Science business (or
rather: I'm in AI for the business, and in CogSci for the science...).


-- Mark

Reply