Dear Matthew, Ed, and Pat, (01)
I agree with the responses by Matthew and Ed to Pat C.
In particular, I'd emphasize the following points: (02)
PC
> The part that I don’t agree with is the implication that the
> “primitives” tactic has been tried . Having an “upper ontology”
> does not mean that anyone has attempted to identify all of the
> known semantic primitives. That is, IMHO, the most effective tactic. (03)
MW
> Yes an upper ontology does not equate to a complete set of primitives.
> An upper ontology can reasonably be expected to capture all
> the ontological commitments, but not all the primitives. I’m not
> convinced there even is a finite set of primitives. (04)
Note that Longman's allows arbitrary exceptions to its defining set
of 2000. For example, the definition of 'lion' cites a *picture* : (05)
Longman's learning dictionary:
> a type of large yellow 4-footed animal of the cat family which
> hunts and eats meat, and lives mainly in Africa, the male having
> a thick growth of hair over the head and shoulders -- see picture
> at CAT. (06)
Aristotle, Galen, Linnaeus, Kant, Whewell, Peirce, Wittgenstein, Rosch,
etc., emphasized the need for a *type specimen* (AKA prototype) and a
detailed description and/or drawing for any naturally occurring species
or phenomenon. Aristotle explicitly said that any proposed definition
will need to be changed as new discoveries are made. So did the others. (07)
I'd also recommend the following book, which is based on many years
of work in lexicography and analyzing large corpora: (08)
Hanks, Patrick (2013) Lexical Analysis: Norms and Exploitations,
Cambridge, MA: MIT Press. (09)
EJB
> The point is that there is no shortage of attempts to solve some of
> these problems. What we need is an assessment of their effectiveness,
> and even better, what properties make them effective? (010)
MW
> What I see is that it is all about use and experience. An early
> experience of using a “defined” vocabulary is that some will
> interpret the terms with their own definitions (for example)... (011)
I agree with both. And I'd qualify Ed's point: (012)
EJB
> I’m not sure whether precision, for example, is part of the solution
> or part of the problem. (013)
That depends on what problem you're trying to solve. For NLP, vague
(or at least underspecified) resources such as WordNet of Anna W's
"semantic primes" are much more useful than precise definitions. (014)
At an ACL meeting, I was talking with Ron Kaplan from PowerSet
(originally from Xerox PARC and later Microsoft). He said that
they had access to full Cyc with all its definitions and software.
But the only part they found useful for analyzing English was the
type hierarchy. They did not use any of the detailed axioms. (015)
For sharing data, such as names and addresses, the very much
underspecified Schema.org has been much more widely used than any
detailed ontology. And one of the chief designers of Schema.org
is Guha, who had been the associate director of Cyc and the original
designer (with Tim Bray) of RDF. (016)
But if you need to do detailed analysis and reasoning (for example,
in designing an airplane, a bridge, or a oil refinery), you need
very precise and detailed specifications. (017)
John (018)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (019)
|