[sc34wg3] TMCL 6.7 Overlap Declaration

Steve Pepper pepper.steve at gmail.com
Sat Jun 20 06:42:29 EDT 2009

* Lars Marius Garshol
| * Patrick Durusau
| >
| > Particularly given the example of person as a topic type overlapping
| > employee.
| >
| > Or,
| > 
| > person overlapping author, editor, artist, illustrator, umpire,
| > referee, chair, secretary, policeman, firefighter, driver,
| > jockey, batter, pitcher, catcher, player, singer, musician,
| > actor, actress, director, producer, composer, etc.
| In my humble opinion, only one of these is a topic type (person).

One day I really must get around to writing up why I disagree so strongly with
this statement, but now is not the time.

Instead I want to support Lars Marius' contention that "disjointness" is the
norm. (It ought to be fairly easy to prove this by querying a representative set
of topic maps, including the more promiscuous ones, like mine.)

Thus, if we agree with Patrick:

| > Making the general rule the most common case seems like a better
| > option to me.

Then things should be left as they are in terms of disjoint classes being the

However, having said that...

The most common exception to the general rule, in my experience, is when people
are classified by profession - exactly the example Patrick mentions.

Arguments about "natural kinds" notwithstanding, it is *extremely common* in
daily discourse to classify people in this way. It should be enough to think of
a typical encyclopedia: there is a reason why the entry on Puccini doesn't start
like this:

	Puccini, Giacomo (1858-1924) Italian person...

If it's that common to classify people by profession in normal discourse, it
ought to be acceptable to do so in a topic map. And not only that, it should
also be easy to do so in a topic map.

Now, the "overlap" between person and (say) composer is *not* a problem (in
terms of TMCL) provided composer is stated to be a subtype of person (which, in
my opinion, it should be anyway). This is because TMCL (sensibly) doesn't expect
a subtype to be disjoint from its supertypes. (In fact, it doesn't really make
sense to talk of "overlap" in a case like this.)

The real problem is in the case of people with *multiple* professions - and
there are a *lot* of those: X was a writer and politician, Y is a composer and
conductor, Z is both singer and movie star. Not to mention Richard Burton, who,
according to Wikipedia, was

	an English explorer, translator, writer, soldier, orientalist,
	ethnologist, linguist, poet, hypnotist, fencer and diplomat.

(Looking at that list almost makes me want to agree with Lars Marius and just
classify him as a person, but I resist the temptation!)

The point to be made is that if this kind of overlap is the most common, then
the mechanism provided by TMCL for overriding the default is suboptimal. The
reason for this is that TMCL requires every pair of non-disjoint types to be
specified separately. This will be unworkable in many cases.

For example, the topic type Instrumentalist is divided into 20 subtypes (from
Bassist to Violinist) in the topic map Tapsa's Concerts (see
http://www.ontopedia.net/omnigator). In order to allow for the fact that any
given player might play two different instruments it would be necessary to make
20! (factorial 20) TMCL declarations: one for each possible pair. That is
2,432,902,008,176,640,000 declarations!

It would be much nicer to be able to say (in one TMCL declaration) that subtypes
of Instrumentalist are not disjoint.

Can this be added?


PSI: http://psi.ontopedia.net/Steve_Pepper
Blog: http://topicmaps.wordpress.com

More information about the sc34wg3 mailing list