RE: SUO: RE: Topic Maps
Matthew,
As I said in another note, I am not arguing for any particular
notation for logic. What I am arguing for is just logic.
>Topic Maps only provides a structuring and linking capability - precisely
>what KIF doesn't have.
There are many things that KIF doesn't have. I regard it as
the "assembly language" of knowledge representation, and there
is a very big need for many different kinds of higher-level
notations that can and should be built on top of KIF. My own
work on CGs is designed to provide other features that can
highlight structures that can be expressed in KIF, but not
as clearly as in a graph notation.
The "structuring and linking" of Topic Maps is also very
important. But as I said, it can be mapped into logic, and
therefore into KIF, CGs, controlled English, and many other
notations. I would just like to see that mapping made more
explicit (most people might not want to see the mapping, but
it should be available to anyone who wants to see it).
Another notation, which I like very much for expressing
procedural knowledge, is Petri nets (a watered-down version
of which is called "activity nets" in UML). In my KR book,
I showed how Petri nets could be translated to logic, in either
KIF or CG notation. That translation, which was much less
readable than the original P-net form, demonstrated very
clearly why multiple notations are extremely valuable for many
purposes.
And by the way, my major complaint about UML (which I also
mention in my KR book) is that
1. UML is a collection of notations that are definable on top
of logic, but the UML developers are afraid that if they
emphasize that point they might scare away people who think
that they don't like logic.
2. They limit themselves to only 6 different notations. I
would prefer to see a more general version of UML, which
was firmly grounded in logic and which supported an open
ended (possibly infinite number) of different notations,
all of which would map to a common underlying logic.
>1. Topic Maps can help us to organise our KIF theories - manage the network
>for us. All we get with KIF is something monolithic.
>
>2. If we are going to get the SUO used, we need to make it easily
>accessible. We could aid this with the use of Topic Maps.
I have no quarrel with those points.
>What is being suggested has nothing to do with how the ontology is defined.
>There is an issue that KIF is unintelligible to all but a few cognoscenti,
>but that is not what is addressed here.
I completely agree. That is why I strongly recommended
the development of some version of controlled English (and
other NLs) as a higher-level, more readable supplement to KIF.
John Sowa