On Dec 13, 2007, at 10:49 AM, Reul, Q. H. wrote:
For these vocabulary to be fully across multiple applications (as intended), I believe that semantic relations can't be either "transitive" or "non-transitive". Otherwise, the working group should give proper guidelines on how to deal with such occurrences.
My reading of Z39.19-2005 §8.3 is that hierarchical relationships in thesauri should always be transitive; §8.4 makes it clear that associative relationships need not be (see e.g. §184.108.40.206).
I've been looking at this issue rather closely lately in the context of the Library of Congress Subject Headings. If BT relationships in the LCSH are treated as if they were, um, Broader Terms, the inferences one obtained are rather idiosyncratic. If SKOS is altered to make broader less-transitive, it would certainly make converting the LCSH to SKOS easier. Unfortunately, it would also make the resulting vocabulary much less useful. It's not SKOS's fault; it's the data.
Incidentally, the problem is less to do with Cows as Mammals, and more to do with Doorbells as Mammals...
Mammals → Primates → Hominids → Human beings → Ethnology → Acculturation → Diﬀusion of innovations → Technology transfer → New products →
Design, Industrial → Engineering design → Structural design → Building → Building ﬁttings → Door ﬁttings → Doorbells