xml-dev - Ontologies, was Re: Schemata are not just constraints

[Jonathan Borden <jborden@mediaone.net>]

Martin Bryan wrote:

... What I am asking for is a clear definition of terms. As Len Bullard
> pointed out, http://www-ksl.stanford.edu/kst/what-is-an-ontology.html
> defines ontology as:
>
> "definitions associate the names of entities in the universe of discourse
> (e.g., classes, relations, functions, or other objects) with
> human-readable text describing what the names mean, and formal
> axioms that constrain the interpretation and well-formed use of
> these terms".
>
> The key phrase in this is "with human-readable text describing what the
> names mean". Without this we are unable to determine anyone's meaning
after
> a few weeks, once we've lost the original context of the message (as
anyone
> returning to the archive for this list server and reading any one of the
> messages in the middle of next year will soon find out). The problem is
that
> neither RDF or Topic Maps have a requirement of the supply of any
> human-readable descripition of the meaning of any referenced subject.
>

    Yet I say that the key phrase in that is "formal axioms that constrain
the interpretation and well-formed use of these terms", otherwise we have a
dictionary and not an ontology.

    Suppose we fashion a similar definition for the source code to a
computer program: "Source code contains human-readable text describing the
function of the code, known as comments, along with a series of tokens
complying to a formal grammar ..."

    True all good source code contains human readable comments as do all
good ontologies, but parsers do not require comments, nor do (ought :-))
comments affect the function of the program.

    Indeed the RDF Schema spec provides for comments and I certainly
advocate comments within any type of schema, yet comments are not the "key"
part of a schema, nor are they the "key" part of an ontology to the extent
that an ontology is designed to facilitate machine processing of
information.

    The real problem is that like good software, good ontologies aren't
trivial to write. Healthcare is a field of endevour highly dependent on
ontologies and has had active development for decades. See the "Unified
Medical Language System" UMLS
http://www.nlm.nih.gov/research/umls/umlsmain.html and see that the
developed onotologies and semantic network serialization formats are not
distributed in either RDF or Topic Map formats (surprise, surprise!). What
is distressing about this is not that XML, nor RDF nor Topic Maps aren't
being used, but rather that many of the ontologies are *proprietary*. The
other issue is one of access. A primary reason to use URIs as a foundation
for RDF based ontologies is that this allows the creation of a web
addressable distributed ontology, and to the extent that information on the
web is less proprietary than for example printed information, such
ontologies might also be less proprietary.

my 2p

Jonathan Borden
The Open Healthcare Group
http://www.openhealth.org