Re: SUO: OpenCyc Motion Open for Discussions
Eric and Chris,
There are many complex issues about how words
are related to word senses, and how word senses
are related to formalized representations in
predicates, concept types, relation types, and
the axioms that define them.
Although I usually agree with Chris Menzel about
technical issues in logic, I think that Chris is
taking too narrow a view about how language,
logic, and ontology are related:
CM> How does a "lexical" axiom differ from an
> "ontological" axiom? Doesn't a lexical axiom
> -- indeed, any nontrivial axiom -- "encode meaning"?
> If so, it belongs in an ontology, according to
> your characterization above, and we've lost the
> distinction between lexical and ontological axioms.
There are very serious differences between the way
words are used in natural langauges and the way
logicians write axioms that define predicates.
I wouldn't use Eric Peterson's term "orthogonal",
but the mapping is certainly not one-to-one:
EP> But isn't the proliferation of word senses
> and micro-senses an artifact of the lexicon
> and somewhat orthogonal to the structure of
> the ontologies in a lattice?
The proliferation of word senses and microsenses
is a fact of life that results from the way people
use language to talk about the world, their thoughts
about it, and their hopes, desires, needs, plans,
and whatever else is humanly conceivable.
Lexicographers do their best to classify word
senses according to the best developed historical
and discriptive practices. There is a considerable
amount of interpretation that inevitably occurs,
but by and large, lexicography is an empirical
practice.
Ontology, however, is far more chaotic than
lexicography. There is a long history of philosophy,
which for more than 2000 years was dominated by
Aristotle's categories. But since the time of Kant,
views on ontology have either proliferated or
fragmented into many different points of view.
In computational circles, you have as many points
of view as there are in philosophy plus a lot of
"seat of the pants" approaches to sitting down
and writing axioms. You also have many people
with backgrounds in lexicography, library science,
linguistics, psychology, and other fields
applying their practices to computational
approaches. The net result can be roughly
characterized as -- pick your preferred term --
chaos, confusion, partisanship, competing
perspectives, etc.
For my recommendation on how to organize the
mappings from words to concepts to theories,
see the paper on "Task-Oriented Semantic
Interpretation":
http://www.jfsowa.com/pubs/tosi.htm
The attached diagram, pointers.gif, is Figure 5
in the tosi.htm article. Athough I admit that
it is still an oversimplification, it is an
attempt to show the complex mappings from
words to word senses to patterns, which I call
_canonical graphs_, and then to theories in the
infinite lattice.
John Sowa
