Bill, (01)
>...An earlier note of yours mentioned "classification". I suspect
>this is how you, being a machine learning person, view the role of
>"ontologies". (02)
I think that may be the first time I've been classified as "a machine
learning person." :-) (03)
As for my view of the role of ontologies, I posted a note on this
forum a couple of weeks ago a note discussing uses of ontologies.
That note didn't mention classification. (04)
>Some quick questions, then. If "ontologies" are supposed to include
>sophisticated mechanisms for handling probability distributions (05)
I think ontologies should be able to include information about
probabilities, and this ability needs to be supported by ontology
languages. I'm not sure what "sophisticated mechanisms for handling
probability distributions" means. Ontologies don't "handle" things.
They represent things. (06)
>and are supposed to support classification, (07)
I'm not sure what you mean by "support classification." Any ontology
that includes a type hierarchy "supports classification" in the sense
that it specifies the types into which objects can be classified and
the attributes that can be used by a classifier. (08)
Adding the ability to express probability information to an ontology
would improve its ability to "support classification." But
ontologies are not classifiers. The primary purpose of most
ontologies is not to "support classification." (09)
>then what would be the role of epistemology and phenomenology in
>such ontologies? (010)
I'm not sure what you are asking here. (011)
It seems when I say ontologies ought to be able to include
probabilities that some people automatically think I'm confusing
ontology and epistemology. (012)
Do you think the Dirac probability rule in quantum theory is
epistemology and not ontology? What about inheritance probabilities
in genetics? What about the sensitivity and specificity of a very
well-characterized diagnostic test for a medical condition? What
about the detection probabilities for a radar system? Emission
probabilities for a radioactive substance? (013)
These kinds of probabilities are well-grounded in physics. I will
grant that there are strict subjectivists who maintain that all
probabilities are subjective. But for some phenomena, probability
distributions are grounded in physical law and are as
well-established and well-characterized as anything we know about the
world. As of the early 20th century, the laws of physics are
fundamentally probabilistic at their core. (014)
I find it very difficult to swallow that the type hierarchy for the
parts in Company X's parts catalog -- something defined by convention
and that can change any time Company X decides to reorganize its
parts catalog -- is ontological, whereas radioactive emission
probabilities are relegated to epistemology. (015)
>Do they all get rolled in? (016)
An ontology represents the types of things that exist in a domain of
application, the attributes entities can have, and the relationships
in which they can participate. I think that when relationships are
probabilistic, and the probabilities are well-established and
well-characterized aspects of the domain, then those probabilities
belong in a domain ontology. (017)
I think that purely subjective probabilities that are grounded in
nothing other than someone's opinion do not belong in an ontology
intended for shared use by a community. Of course, if somebody
chooses to use a probabilistic ontology language to formalize his
personal probabilistic theory of a domain, no one can stop him, just
as they can't stop him from writing an idiosyncratic OWL ontology to
which nobody but he subscribes. In fact, the languages developed for
the purpose of formalizing shared domain knowledge may turn out to be
very useful for formalizing the opinions of individuals. But I would
use the label "ontology" to mean a repository for well-characterized
domain knowledge agreed upon by a defined community and shared widely
among practitioners of the community. I think the kinds of
probabilities I described above pass that test. (018)
>Is it the case that you reject the traditional divides between
>ontology, epistemology, phenomenology, and semantics in natural
>languages? (019)
No, I do not reject those distinctions. (020)
I should point out, though, that ontologies are intended to represent
what "really is out there", but they are not THE SAME AS "what is
really out there." The very fact that we do not agree on our
ontologies is proof that our ontologies contain representations that
are not fully faithful to what they represent. Analysis of the
relationship between a formal ontology and that which it represents
belongs to the field of epistemology. Right? (021)
Kathy (022)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: 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 Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (023)
|