About the model: Question 3
3. Would I pay a price for defining expression as a subclass of work, manifestation as a subclass of expression, item as a subclass of manifestation, etc.?
“Always ; the better you define anything, the more chance that definition will break in some other context. I know this sounds terribly gloomy, but it’s just a fact of life that when I say “that is a zebra” (which most people will be able to identify just fine) some weird zoologist will come along and ask “what sort of zebra?” (yes, there’s more than one, and the differences between them cause great problems within zoological taxonomies and is mostly unresolved)”–September 11, 2007 email from Alexander Johannesen.
The RDF primer in section 5.1 says that any instance of a subclass is also an instance of the class. I think it is correct to say that any item is also a manifestation, expression and work. Am I misinterpreting anything here?
“No. But I do fear that the taxomatic constraints of class-instance hierarchies are somewhat flawed the further away from the original definition one goes. Some talk about ontological distance (how many jumps from where you are now to the thing that originated the classification path) needing to be weakened to the proportion of each sub-class’s weight (which, funnily, is a recursive argument and can’t be solved, certainly not by humans). It basically mean that you need to have full knowledge of what you’re about to classify before you can classify the thing. Doesn’t work, does it? 🙂RDF and Topic Maps are at best “close enough” for the things we’re trying to do, although my personal view is that all of this modeling business is followed as it was the truth because it looks like the truth. I’m no longer sure we’re doing what’s right, but I digress.Back to your question here ; a thing that’s an instance of a thing is by inheritance an instance of all parent classes. For example, if we use this silly classification tree ; thing living thing walking living thing humanWhat’s meant is that an instance of human is at the same time a thing, a living thing *and* a walking living thing. Things are whatever is up the tree, and this lies at the root of taxonomical classification theory. For a thing to be anything, it must be what has passed before it. That’s the tree of knowledge for instances of things, but I’m sure I’m preaching to the choir on this one. 🙂
In your problem, the thing that precedes it up the tree isn’t really an instance of that thing, so here I wouldn’t use instances of things, but properties. But then again, neither of these models are satisfactory in my eyes. This is where I suspect RDF:OWL comes to the rescue where you can make further qualifications to the instances to relieve them from parent class inheritance, but I’d have to read up on it (been a while)”–September 11, 2007 email from Alexander Johannesen.
I note that there has been at least one attempt to express FRBR in RDF and that they have not defined expression as a subclass of work. Instead, they have defined work as “disjoint with” expression, manifestation and item. Yet the FRBR primer (section 5.5), following OWL, says that when two classes are disjoint “no resource is an instance of both classes.” If an item is also a manifestation, expression and work, the FRBR-RDF people are wrong to do this, aren’t they?
“Ah, yes, I should have read a bit more of your email to find what I was trying to say with the previous section. 🙂 I think it would be worthwhile to ask them why they cose to do so ; I susect there’s limitations between the taxonomical hierarchy and how OWL operates on atomic entities that share properties but not class instantiations. (Hm, that sounded Greek to me. Let me know if this comes across as such.)”–September 11, 2007 email from Alexander Johannesen.
Gordon Dunsire (email to RDA list) refers to FRBR having a “no manifestation without expression rule.” DECISION: Bruce et al. have persuaded me that something should be defined as a subclass only if it fits into an “is a kind of” sentence. Music is a kind of work, but expression is NOT a kind of work. Therefore, I am following the FRBR modelers and using the OWL disjoint with convention after all. They assure us that this does not prevent an expression from inheriting subject headings from a work, or affect the ability to assemble all the expressions of a work for a user, or to identify a manifestation using its work identification information or its expression identification information, as long as the relationships (manifestation to expression and expression to work) are established.