[topicmapmail] Opera topic map question

[Murray Altheim]

[this sat in the moderators in-tray for a few days; resending to the
list from my older email address...]

Alexander Mikhailian wrote:
[...]
> Actually, I browsed already through the Italian Opera topic map and
> here is one question.
>
> * Writers, composers and librettists do not depend on persons.
> However, in the context of this precise topic map, I would find it
> useful to see a supertype-subtype hierarchy under "person".
>
> * Dates do not form a hierarchy either, although I would naturally
> expect "date-of-birth" be the subtype of "date" also.
> In my opinion, the only reason for not creating a hierarchy for these
> two topic sets is because  these topics are referenced by
> PSIs and a hierarchy would contain assumptions about the relations
> between topics referred by PSIs that are not true outside this
> precise topic map.
>
> Since the Italian Opera topic map is an exemplary document, I believe
> it's ontology has been thoroughly thought and the possible
> hierarchies are not there for a reason ;-)

Alexander,

A common problem with computer-based ontologies is underspecification,
a lack of precision in the underlying model, or perhaps a lack of any
formal model whatsoever (despite descriptive prose and/or appearances
that might indicate otherwise). There are many different (and sometimes
largely incompatible) foundations for what people informally call a
"hierarchy", including a basis in "sets", "classes", or "collections",
each with their own underlying logics/models. Some models are not based
on a formal (i.e., mathematical) "logic" at all. That's okay too.

Because humans do not typically converse using formal, precise, logical
language, we tend to overlook these issues until our lack of precision
elucidates an obvious fallacy. The effect of this may be inconsequential
or something much more serious, depending upon the nature and scope of
the application. If Topic Maps were being used as DARPA is using RDF-
based technologies, people might even be erroneously characterised due
to faulty inferencing, the consequences of this for them may be rather
dire.

A seeming alternative to such misuse of logic is posited by various
neo-pragmatists of the Dewey/Sellars/Rorty school (as well as Habermas
if I'm not stretching this too far), who advocate the view that such
uses do not constitute expressions of logic but rather of human
communication, what Wittgenstein termed a "language game" (which
unfortunately and unintentionally may connote frivolity when it is
really intended to denote that language has no inherent meaning apart
from its contextualised usage as a human activity).

My point here is that a Topic Map should not be considered more precise
than the use of natural language that forms its constructs, and indeed,
implying that a system based on informal language has a more formal,
logical underpinning greatly increases the likelihood that there will
be significant errors in its interpretation (using the term loosely
here) by either machines, human programmers or human users of products
generated from those Topic Maps, based on inherently faulty assumptions.
As a participant in the design of XTM 1.0 I'm actually very comfortable
with the idea that we left those two association types loosely defined.

It's probably better to design into the systems themselves (as well as
noting in their documentation) a cautionary warning indicating that the
systemic use of natural language is distinctly not formalised rather
than pretend otherwise.

Of course, the above opinions are my own and I should caution that this
is not the typical view held by most practitioners I am aware of in the
Knowledge Representation (academic) community.

Murray

PS. as a reply to your recent message regarding the two "association
templates" in XTM 1.0, it might be worth noting that they were rather
deliberately underspecified because we (a) did not base XTM 1.0 on a
formal mathematical model nor wish to imply that we were, and (b) we
couldn't at that time agree on a means of specifying "association
templates", partly because this whole territory is a bit thorny -- as
you have correctly gathered -- hence the entire basis of those two
association types was described in a relatively informal prose, not
in logic. I can't answer your final post script question, but my
suspicion is that unless XTM 2.0 *explicitly* states an identity with
XTM 1.0 as regards those two association types (and their roles) then
one should never assume such an identity, even if the language is
similar or the same, as the models of XTM 1.0 and 2.0 are different.
...........................................................................
Murray Altheim <murray07 at altheim.com>                           ===  = =
http://www.altheim.com/murray/                                     = =  ===
SGML Grease Monkey, Banjo Player, Wantanabe Zen Monk               = =  = =

       Boundless wind and moon - the eye within eyes,
       Inexhaustible heaven and earth - the light beyond light,
       The willow dark, the flower bright - ten thousand houses,
       Knock at any door - there's one who will respond.
                                       -- The Blue Cliff Record


-- 

...........................................................................
Murray Altheim <murray07 at altheim.com>                           ===  = =
http://www.altheim.com/murray/                                     = =  ===
SGML Grease Monkey, Banjo Player, Wantanabe Zen Monk               = =  = =

       Boundless wind and moon - the eye within eyes,
       Inexhaustible heaven and earth - the light beyond light,
       The willow dark, the flower bright - ten thousand houses,
       Knock at any door - there's one who will respond.
                                       -- The Blue Cliff Record