[topicmapmail] OKS 2.0 & Omnigator 007

[Kal Ahmed]

On Sun, 2003-12-28 at 20:45, Murray Altheim wrote:
> Lars Marius Garshol wrote:
> [...]
> > Kal has used a natural language approach that basically appeals to
> > intuition. That's pragmatic, but I guess you are right that it might
> > not be as precise as it could be. Whether that will be a problem I
> > don't know.
> 
> I don't know why I should be having any difficulty with this, now
> that I understand it better. If you think about my application,
> Ceryle, its intended audience is authors, not ontological engineers,
> so I've been dealing with this kind of "intuitive" approach to
> ontology-building/viewing for about two years. I *hope* it won't
> be a problem. :-)
> 

There are also UML diagrams for each of the patterns I have proposed. I
feel that for the purpose of pattern description, that is about as
formal as one should get in the description. Of course, patterns can
give rise to more formal expressions in specific syntaxes (e.g. TMCL
statements). But the target audience for patterns (as I see it) are not
interested in learning yet another modelling language, but instead in a
prose description and a diagram (preferably in an already-understood
notation). Of course, I might be wrong ;-)


> > | What I'd like is to be able to create that relation hierarchy and be
> > | able to know for certain for any specific relation type (for a
> > | specific hierarchy type) whether or not various types of inferences
> > | are available. 
> > 
> > That's a valid point. I think Kal's PSIs as defined do not really
> > license any inferences.
> 
> No, I'd agree that they don't, partly because the kinds of instances
> of association types and role types they are the type for have a
> variety of kinds of inferences (or in some cases probably none).
> 

I'd agree too. Most of the patterns proposed so far are patterns for
organising topics - and I'm not really sure that there is much to infer
at all.

> > | Display is one thing, but it's (for my purposes, anyway) more
> > | important to have some handle on what can be done with those
> > | relationships, e.g., being able to answer questions such as is a
> > | "customer" a kind of person or a kind of company, or both? Or
> > | similar examples in zoological taxonomy, natural language
> > | processing, etc.
> > 
> > If that's what you are looking for then these PSIs won't provide it,
> > and I guess you should look elsewhere. These PSIs were really meant as
> > a kind of semantic annotation for humans (or smart machines), and not
> > to provide a basis for inference.
> 
> It occurs to me to look at the fact that I'd developed some PSIs
> that do basically what Kal's do, in that I early on needed a way
> to create "association templates" to tell my graph visualizer
> which way to point the arrows in the displayed edges/arcs. I have
> developed search and navigation features based on the hierarchy,
> not on the superclass/subclass associations (though they are
> templated with the hierarchy PSIs so they have direction too). So
> all I really need to do is add the "subordinate" and superordinate"
> PSIs to the subject identity for my PSIs and we're all copacetic.

Yep, that sounds like a match to me.

Cheers,

Kal
-- 
Kal Ahmed, Techquila
Standards-based Information Management
e: kal@techquila.com
w: www.techquila.com
p: +44 7968 529531