[Corpora-List] WordNet vs Ontology

[John F Sowa]

Many of the comments in this thread require some clarifications.

Yannick Versley
> An ontology is, intuitively, a collection of facts about a set of entities.

That is too weak a definition.  A database is a collection of facts.
But an ontology has a normative effect:  It specifies definitions
and axioms that have been agreed to as a foundation for a family
of interoperable databases and knowledge bases.

YV
> It may have semantic relations between concepts, as in DBPedia,
> ConceptNet or the various "extended" wordnets.  This is the
> prototypical case for people from the "semantic web" community

I agree.  But note that those collections are more firmly established
than just an arbitrary collection of facts.  They are intended for
some larger group to use as a normative basis for organizing their
databases and knowledge bases.

Robert Amsler
> WordNet is not an ontology. Wordnet is a tangled hierarchy. A taxonomy
> that allows multiple upward paths from a given node.

I agree with the first two sentences.  But multiple inheritance
(i.e., multiple upward paths) is a requirement for a lattice.
Any formal ontology that is not represented as a lattice is usually
incomplete -- it ignores (deliberately or accidentally) some options
that are implied by the axioms and definitions.

Lattices have been widely used in computational linguistics since
the early work at CLRU in the 1960s.  For a brief overview of lattices
in ontology from Leibniz to the facet analysis by Ranganathan to the
widely used methods of Formal Concept Analysis (FCA) see slides 94 to
98 of http://www.jfsowa.com/talks/kdptut.pdf .

To see how FCA software can derive lattices from Roget's Thesaurus
and WordNet, see slide 104.  You can follow the pointers to the FCA
demos.  Type any English word to those demos, and you'll see a lattice
of the "concept neighborhood" for that word, as derived automatically
from wordNet or Roget's.  WordNet, by the way, has very poor coverage
of adjectives.  Roget's Thesaurus is a useful supplement to WordNet
for adjectives and adverbs.

Jim Fidelholtz
> While, as a mathematician and sometime dabbler in philosophy,
> I am familiar with logical notions, as a linguist I have almost
> always been very skeptical of the applications of logic to
> linguistic structure. Many very brilliant people have attempted
> to force language into a logical mold, to my mind with *very*
> limited success.

I very strongly agree.

But I also make the point that language is sufficiently flexible
to be used with any degree of precision or vagueness that is
appropriate to the subject.  Two scientists, engineers, or
mathematicians can talk on the telephone and get their point
across in their native language just as precisely as they could
when writing formulas on a blackboard.

The source of the vagueness is not language. It's the subject matter.
You can't talk precisely about a vague subject.  But two experts
on any precisely defined subject -- chess, bridge, or Unix --
can talk about it precisely over the telephone.

John

_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora