[sc34wg3] SAM-issue term-scope-def
Bernard Vatant
sc34wg3@isotopicmaps.org
Thu, 13 Jun 2002 08:53:53 +0200
Hello all.
Lars Marius asked me on another forum to bite here. Well, I was very uneasy with that
issue until yesterday night, then I got the following idea(s), which maybe could help.
Sorry, it's a bit long. But I tried to be very explicit. Thanks for your time and
attention.
I've figured so far scope as a bag where you could put so many different things that no
general rule how to process it could be really set, other than arbitrary or "less evil",
as Lars Marius tried to figure below . But in fact there is a way to entangle that, if you
begin by expliciting the implicit (meta)association expressed by scope. The role-players
of this (meta)association are:
1. The scoped assertion itself (reified)
2. The elements of the scope set (scoping topics)
Note that "assertion" in 1 is to be understood in the most generic sense: any assertion
that can be scoped, including regular associations, topic-name and topic-occurrence
associations.
If the role of the first member (assertion) is clearly specified (scoped), the role of the
scoping members are not specified, further than "scoping" ... which is a very poor
semantic indeed. That's where the issue comes from. If you specify, like in any regular
association, the role of scoping topics, things become clearer and the conclusion is
easier to draw and quite natural.
Let's take an example. "Henry is King of Navarre and King of France from 1589 to 1610"
Express that using the basic assertion "Henry is King", the rest of information being
expressed as scopes. Note BTW that this assertion can be expressed like a regular
association between Henri (person) and a status (King), or the attachment of a name
(Henri) to a topic (King) ... and that it does not matter in fact. In any case, we have
three scoping elements : France, Navarre, and the time span.
Clearly, the two former have a "space" role, and the latter a "time" role in the scoping
meta-association. So if you express it explicitly, quite obviously you have the following
members:
{Henri is King} (scoped assertion)
{France, Navarre} (space scope)
{1589-1610} (time scope)
Now it figures that scopes playing the same role are first gathered by a "OR" (union?),
and resulting sets (members) are gathered by a "AND" (intersection?) to get the global
scope. More precisely:
"Henri is King" is valid if ("space belongs to Navarre" OR "space belongs to France") AND
"time belongs to 1589-1610".
It figures another thing. Asserting a scope seems to be in that case setting implicitly
the domain of two variables, "space" and "time". The nature of those variables are
expressed by the role specifications in the meta-association.
If we try to draw a general rule for that:
If : the scope set is S = {Ai, Bi, ..., Xi, ...}
where A, B, ..., X are scoping variables and Ai, Bi, ..., Xi values of those variables.
Then : the scoped assertion is valid if
AND( A belongs to {Ai} , B belongs to {Bi} ... X belongs to {Xi} )
It seems that this is very generic, the scoping variable-role, besides time and space,
could be language, source, thesaurus, market sector ... whatever. Of course, this rule can
be applied if roles of scoping elements are specified. Maybe the standard should be
extended to include that? From a syntax viewpoint, it would mean allowing <member> as
child of <scope> ... Sorry to speak syntax here :))
What about now unconstrained scope and merging? The original source(s) of merging being
added to scopes as reified topic maps in the merged map, as I suggested in a previous
post, seems consistent with that. The original topic map plays role "source". In the above
example, you could add to scope that the assertion (unauthorized) source is "TM-BV".
So, after merging, you would have another member in the scoping meta-association:
{Henri is King} (scoped assertion)
{France, Navarre} (space scope)
{1589-1610} (time scope)
{BV} (source scope)
If you have the same assertion confirmed by "Larousse 1972" (just checked), this will be
added to the source member:
{Henri is King} (scoped assertion)
{France, Navarre} (space scope)
{1589-1610} (time scope)
{BV, Larousse} (source scope)
Well. I think that makes it for my 0.02 Euros.
Bernard
-------------------------------------------------------------------
Bernard Vatant
Consultant - Mondeca
www.mondeca.com
Chair - OASIS TM PubSubj Technical Committee
www.oasis-open.org/committees/tm-pubsubj/
-------------------------------------------------------------------
----- Message d'origine -----
De : "Lars Marius Garshol" <larsga@garshol.priv.no>
À : <sc34wg3@isotopicmaps.org>
Envoyé : dimanche 9 juin 2002 19:45
Objet : [sc34wg3] SAM-issue term-scope-def
>
> In ISO 13250 the definition of scope says that scope is the union of
> the themes. That is,
>
> [finland = "Finland" / norwegian swedish]
>
> means that the name "Finland" is valid in Norwegian and that it is
> also valid in Swedish.
>
> XTM 1.0, on the other hand says that it is up to the application how
> to interpret the scope.
>
> The current SAM version[1] says scope is the intersection of the
> themes, which means that this particular scope would only be valid in
> *both* Norwegian *and* Swedish at the same time, which would require
> the topic to be written in this way:
>
> [finland = "Finland" / norwegian
> = "Finland" / swedish]
>
> On the other hand, that something is my opinion today could according
> to the SAM be expressed by scoping it with topics representing me and
> today. ISO 13250 explains in a note that to say this would require a
> topic representing me today would have to be created and used as the
> scope. (No mechanism is provided for doing so in ISO 13250, however.)
>
>
> The question is, which of these three approaches should we choose? I
> think we need to consider carefully the consequences for the
> unconstrained scope, for merging, and also for querying operations.
>
>
> --- Scope is union
>
> If we say that scope is union that would mean that the following topic
>
> [finland = "Finland" / norwegian swedish
> = "Finland" / norwegian
> = "Finland" / swedish]
>
> contains two redundancies, and is actually equivalent to
>
> [finland = "Finland" / norwegian swedish]
>
> which would seem to imply that the rules for equivalence and
> redundancy elimination need to be modified.
>
> Furthermore, if associations with published subjects are necessary to
> form compound scope it means that in order to answer the question
> "what occurrences of the topic 'term-scope-def' are valid in the scope
> 'lmg'" correctly when presented with
>
> {term-scope-def, opinion, ""} / lmg-today
>
> the topic map processor would have to traverse associations to
> evaluate the scope correctly, and building compound scopes becomes
> very cumbersome.
>
>
> --- Scope is intersection
>
> This means that to assert that something is valid in two different
> contexts requires it to be repeated. This seems like the lesser evil.
> It also means that the arithmetic of scope works well with the
> unconstrained scope being the empty set. (See previous email.)
>
>
> --- Scope is unspecified
>
> One of the mechanisms for merging (TNC merging) would then depend on a
> feature whose meaning is unspecified. Furthermore, when attempting to
> merge topic maps that interpreted scope differently one would be bound
> to get suboptimal results. Furthermore, we would have to create TMQL
> scope operators for a construct whose interpretation is undefined.
>
> In short, this does not seem to work very well.
>
>
> [1] Versions of SAM prior to 1.23 had an editorial error in the
> definition of scope.
>
> --
> Lars Marius Garshol, Ontopian <URL: http://www.ontopia.net >
> ISO SC34/WG3, OASIS GeoLang TC <URL: http://www.garshol.priv.no >
>
> _______________________________________________
> sc34wg3 mailing list
> sc34wg3@isotopicmaps.org
> http://www.isotopicmaps.org/mailman/listinfo/sc34wg3
>