xml-dev - Re: local, global (was various ontology, RDF, topic maps)

[Jonathan Borden <jborden@mediaone.net>]

Simon St.Laurent wrote:

> In all of this recent talk of creating ontologies, using schemas for
> constraints, and creating large-scale distributed networks of commonly
> understood information, I feel like we're seeing the usual formula of
> 'achieve agreement, implement everything according to that agreement,
enjoy
> paradise'.
>

    This has been the problem of structured medical terminology for the past
3 decades and today we have no universal agreement about and less structure
in medical records. And then the government steps in and says "We aren't
requiring you to do it this way but if you don't you won't get paid." That
is the best way I know of achieving consensus. My proposition is, simply
put, that while we are cleaning up the medical claims procedures, we seize
the opportunity to fix the information system.
...
>
> Developing standards which work locally seems like enough of a challenge,
> and developing standards which work globally seems like a project better
> left for future development, after we've figured out what might make sense
> in the less costly though perhaps less-inspiring world of local
> communications and understandings.  Perhaps that way we can dwell less
upon
> the credentials of those who create our toolkits and more upon the task of
> creating our projects ourselves.
>

    Again the principle of reification, or in plain English the ability to
make assertions regarding the classifications allows us to place these
ontologies into "contexts" or "spaces". One's regard for a particular
ontology is then applied to its context. As an individual, I can use only
those ontologies in a particular "context" and I can apply a belief value
(i.e. create a Bayesian network) to individual statements or entire
ontologies as I choose.

    I might employ a suitable "agent" to process my own information with
fuzzy logic (for example). You might blindly believe the network (your
choice). The benefit to creating standard *syntax* for defining ontologies
is that each of our agents can choose its own processing model, yet operate
on the same data.

    Technically a "context" or "space" may be represented in RDF using a Bag
(a collection), which is simply a resource node. "Membership" in the context
(or containment in the space) is denoted in the following fashion:

    <Description bagID="Context-1">
        <foo ID="S1">1</foo>
        <bar ID="S2">2</bar>
        <baz ID="S3">3</baz>
    </Description>

    which is parsed into the RDF triples:

    [Context-1, _1, S1]
    [Context-1, _2, S2]
    [Context-1, _3, S3]

and

    [S1, foo, 1]
    [S2, bar, 2]
    [S3, baz, 3]

One can now attach a belief to the entire collection:

S4:    [Simon, believes, Context-1]
S5:    [S4, beliefValue, 0.66]

Again, think of RDF as an assembly language for semantic reasoning, but
realize that real benefits won't be achieved until we get high level
languages and tools.

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