Re: [topicmaps-comment] TAO vs. ERA

["Thomas B. Passin" <tpassin@home.com>]

[Lars Marius Garshol]


>
> * Lars Marius Garshol
> |
> | I do have a problem with your process for arriving at the 8 topics
> | relevant to "airplane", though. You say you select the scoping topic
> | "airplane", and this somehow produces the list of 8. By scope
> | filtering? If so, how?
>
> * Thomas B. Passin
> |
> | You retrieve only topics whose name has a scope that matches the
> | scoping topic named "airplane".
>
> Right. So to find the topics relevant to "airplane" you find all
> topics that have names in that scope. That sounds like an abuse of
> topic maps semantics to me, for two reasons. The first is that if the
> topic has "airplane" in the scope of one of its names that must mean
> that the name is somehow restricted to that scope, which I doubt is
> the case. The second is that it does not imply in any way that the
> topic is related to airplanes.
>


This is an example why semantic modeling is so hard.  I think it is a very
appropriate use of the semantics of scopes.  Consider a famous airplane, the
Boeing B-17.  A common name for it was "Flying Fortress".  Boeing would have
had a number of model designations for it.   The military called various
models "B-17A", "B-17G", etc.

I could consider this thing in terms of a symbol, a warfighting strategy,
military history, the development of the field of military aviation, methods
of aircraft manufacture, as an airplane, etc.  Perhaps I want different
names to come up when I consider military aviation than when I want to
consider aircraft manufacture.  Even if I have only one such scope at some
particular time, that does not mean it's inappropriate to have it.

Now I could certainly create an association as you say.  There are
alternative approches to most things in topic maps.  In fact, you could make
scopes into associations (I suppose you would have to give names reifying
topics to do so).  But it seems to me that the way I've described is
precisely in line with the use of scopes as described in the various specs
and recs.  I also think that is is convenient and appropriate.

There is a (strong) need to provide a context for subjects (I'm speaking in
human terms, no computer science terms here), and associations aren't for
that, they are for asserting relations between topics.  Scopes are ideal for
providing contexts.  I consider this to be one of the strengths of topic
maps.

In terms of the above example, let's say I have an association as you
suggest, and I also have other associations for the other areas of interest,
that invilve some of the same topics.  I'm not saying that would be bad, I
might very well do it.  But now I want to find which ones of those involve,
say, the military.  The best way for me to find them will be if they (the
associations) are tagged with a scope of "military".  Now, is there really
any different between this application of scopes and using them to pick out
French vs. English labels?  I don't so.

I think the explanation of what a "scope" is supposed to be is weak.  I've
said that before.  Here we see that weakness is making itself known.  But
scope is the part of topic maps that is the least familiar.  People know
about relationships between things - entities - topics - concepts from many
places, people know about names, people know about roles and instances.  But
you won't find scopes in standard KR systems (say conceptual graphs, KIF,
etc.).  So scope ought to have the strongest explanation and definition of
semantics of anything in topic maps, but it doesn't.

The XTM Rec does use the term "context" about scopes:

"It establishes the context in which a name or an occurrence is assigned to
a given topic, and the context in which topics are related through
associations."

It also talks about the "extent of validity" of these assignments.  For me,
"extent of validity" and "context" are not congruent - they have some
different connotations.  This leaves me in doubt about the intent.  Anyway,
"validity"  seems to be some logical term, which would mean we need some
system of logic to define its precise meaning.  I doubt we want to get into
that.  On the other hand, for indexing, bibliography, finding information,
and so on, "context" (in everyday usage) is very valuable or even essential,
so it makes sense to me that we have something specifically intended to
fulfill that need.

In fact, the ISO TM Spec has a subtly different wording - it uses "context"
to ***explain*** what it means by the term "validity".

If we should further clarify just how  scopes are to be used, we should be
flexible and not restrictive.  I'd like to see the part about validity just
fade away - the remaining text about context works very well on its own, and
that would be consistent with the ISO wording.

> I think what you want is a "related-to" association between "airplane"
> and the 8 topics, rather than use scope.

I have found it useful to create such an association and THEN to scope it
with a scoping topic of "airplane".

>Or, actually, the "related-to"
> association is just an untyped association, so you'd most likely want
> to express what the relationship actually is. And then you'd most
> likely find that these 8 topics are terms relevant to the field of
> "aviation", or something like it.
>

According to this approach, how do I find terms "relevant" to the field of
aviation?  Now I have to look at the types of associations, but then what?
With your method, it seems that I need an exhaustive list of which topics
are "relevant" - what is the nature of this list? - and them check the type
of each association against this list.  The type of the association type
won't do it, neither wil the names of the association types.

As best I can see, the only method we have for such a list is scopes, unless
we want to construct a special set of associations to hold such things.  If
we did that, we would have an association instance of type "topics relevant
to a term", which had a role called "relevant term" filled by a topic named
"aviation", and we would have to check the type of our original association
against the role-players of this new association.

This would be extremely complex and idiosyncratic (which is to say, not
usable by many processors), while using the existing scope mechanism is
simple, clean, and could easily be processed by most any topic maps
processor.

These are some of the reasons why I'm using scopes this way.  There are
others...

Cheers,

Tom P