Re: SUO: Re: Parse Of Things Remembered
Pat,
What I was trying to illustrate was very simple and mundane:
>>This example does capture the three-way relationship, but
>>only by creating another entity "giving", which itself has
>>the open slots in its definition -- formally speaking,
>>any representation of "giving" must have "frame-like" or
>>"lambda-calculus-like" representation, which contains
>>three inner variables or slots. Then each of the three
>>sentences instantiates one of those slots.
PH>OK, if THAT is the point, then of course I will agree with it
>immediately: we will need to write axioms that mention more than two
>things, or use more than two relation symbols, or whatever.
That is all I was trying to say. I drew another diagram that
shows that you can't get rid of the three-way junction by
replacing "giving" as a relation with three arguments by
an entity linked to three relations. Following is a little
diagram that shows the two kinds of CGs for "give" and "gives":
http://www.bestweb.net/~sowa/cg/give.gif
The oval labeled "gives" in the top diagram represents a
relation with three arguments. The box labeled "give"
represents a distinct entity with its own existential
quantifier, which is linked to three dyadic relations.
There are many arguments pro and con for one representation
or the other, but the only point I was trying to make is
that both versions have a central node with three links
attached to it.
PH> But now I
>am puzzled by the emphasis on threeness for a different reason, since
>it seems to me for example that if we are to really capture the full
>meaning of 'give' we will probably need to relate it to about 7 or
>more other concepts (whether they are expressed formally as relations
>or names or whatever).
Yes. I would put all those additional concepts (or whatever)
into the definition of "give", along the following lines:
give = (lambda x,y,z)(...expression involving x, y, and z...)
That expression might be arbitrarily complex, but it has three
links that are "exported" via the lambda abstraction.
>If we are at this rather loose level of
>discussion, I don't see how one can insist that "threeness" is
>particularly important. A giving involves two agents, a thing given,
>an intention (maybe several), a time, a place (usually), probably a
>reason, maybe an overarching larger event or circumstance (such as a
>birthday), etc. . Most things have many connections to other things,
>and (depending on how strictly one understands 'full meaning') their
>full meaning cannot be stated without making reference to more than
>three of them. So why do you stop at three?
I agree. I don't want to stop at three, but for the moment,
I am happy that we all agree that there are more than 2.
>A related issue is the idea, which arises in various forms, that any
>particular number of relations is somehow indicative of a particular
>semantic content, as for example the idea that threeness is somehow a
>litmus test for the presence of intentionality.
I would say that the comparison with a litmus test is quite apt.
For one thing, a litmus test only gives you one tiny bit of
information: whether a solution is acidic or basic. The only
information that the threeness test gives you is that you need
more than two arguments.
There are many kinds of concepts that require more than 3
arguments. Intention is just one, betweenness is another.
To distinguish them further, you need more tests. I believe
that we should have large numbers of tests (at least one for
every distinction in the ontology -- which may number in the
thousands).
John