data modelling and FRBR WEMI ontology

Karen Coyle writes on the RDA listserv:

FRBR claims to be based on a “relational” model, as in “relational database.” That is not tomorrow’s data model; it is yesterday’s, although it is a step toward tomorrow’s model. The difficulty is that FRBR was conceived of in the early 1990’s, and completed in the late 1990’s. That makes it about 15 years old.

I think it would have been just as much a mistake to tie the FRBR model to an RDF model as it would have/was to tie it to a relational database model.  Whatever we come up with is going to last us more than 15 years, and things will change again. Now, I’ll admit that I’m heretically still suspicious that an RDF data model will in fact be ‘the future’. But even if it is, there will be another future (or simultaneous futures plural).

I think there are some fundamental principles to data modelling for information processing that are independent of particular technologies.  Now, maybe not completely independent, different technologies have different ‘affordances’, and that, along with changing technological capacities and just increased experience and practice do lead to a constant evolution of best practices and understood principles of data modelling, sure.

But I think they’re still there. I don’t know if there are any good treatises on such a topic, fundamental trans-technological data modelling principles. (Anyone know?). If there are, I wonder if they are very mathematical, formal logic and such. (I strongly suspect that, mathematically, it can be proven that rdbms and linked data and set theory and object oriented models, etc., are all actually mathematically equivalent in expressiveness. Curious if anyone has actually done so. But at any rate, that’s related to the existence of some basic principles independent of technology or serialization, I’d think).

One of the fundamental aspects of data modelling is deciding what the ‘things’ are that you are modelling (“ontology”). And I think the FRBR WEMI ontology gets it pretty good — whether you are using a relational database or linked data or anything else we know of. The FRBR WEMI ontology does a pretty good job of specifying the ‘things’ in the bibliographic/documentary universe that we should model, in order to ‘say’ things about these things to create information systems to serve our users. I’ve tried to explain why before, I will surely try to do so again.

(Whether the particular attributes and relationships added on to the WEMI ontology are as right, I’m less sure. Even the FRBR report is more hesitant and provisional there itself.  Whether RDA’s increased specificty/formalization of the FRBR model is as good, or the RDA linked data vocabulary formalization is as good, I don’t know enough to say.)

I seem to remember hearing that when the FRBR report was being worked on, some people (Svenonius?) wanted it to be expressed in the language of set theory instead. But my impression is that ‘relational’ language won out because the FRBR committee believed that was “the language that computer programmers wanted it in.”  Ironic that that choice has led to the opposite outcome, to some in the more CS/programming/IT oriented community dismissing the FRBR model because it uses relational language.

Soon after the FRBR report came out, some argued that it should have used “Object Oriented modelling” language instead, that ‘relational’ was old, and ‘OO’ was new.  Of course, now, everyone thinks OO is old news, and everyone wants ‘linked data’ (or in other sub-communities, ‘type theory’).

I tend to think they should have just gone with ‘set theory’ oriented language, because it is, I think, the most clear, while still being abstract enough to make it harder to think the WEMI ontology is tied to some particular technology like relational databases OR linked data. I think WEMI gets it right regardless of whether you speak in the language of ‘relational’, ‘set theory’, ‘object orientation’ or ‘linked data’/RDF.

4 thoughts on “data modelling and FRBR WEMI ontology

  1. Now is the time for a pedant to step up & explain how the “relational” in “relational database” does not mean relationships between entities. (See the articles on relational algebra on wikipedia).

    The “entity analysis” that the FRBR group says they did seems to fit pretty well what people do when they are designing ontologies for RDF, so I don’t really see the issue here.

    I’m glad, at least, that the idea of using OO concepts has fallen by the wayside. As for type theory, I cannot even imagine what it would bring.

  2. Thanks Erik, quite so, sorry if I was confusing (or confused!) in conflating rleational databases/relational algebra/entity-relational modelling.

    Ah, type theory. One of the articles Karen suggested I read on why WEMI is not a good ontology for linked data/RDF… well, it kind of swamped me, it was a dense article…. but I _think_ it was actually trying to critique it on ‘type theory’ grounds. I may be completely wrong.

    Something I could write another blog post about though, is how I think that may be related to the fact that, well, _no_ model matches the world exactly, they’re _all_ approximations, and for that how to model something is always context-dependent and subjective. Sometimes I see someone try to bash a _particular_ model on those grounds, but I think it’s pretty clearly true of ALL models. (but more obvious with some than others, and the fairly abstract nature of the ‘bibliographic universe’, that most of the ‘things’ we talk about are NOT in fact single physical objects you can point to, makes it more evident than people who can at least pretend they’re modelling ‘real things’). I think a ‘positivist’ approach to modelling is inherently doomed, I’m ‘constructivist’ to the core when it comes to modelling (and there is in fact plenty of scholarly writing in information science and documentary studies, not to mention plenty of other places, that agrees with me).

  3. Jonathan, you weren’t really confusing, I just wanted to clarify things.

    Yes, modelling the world is always an approximation, and the usefulness of a model depends on who is using the model. I began to realize this after reading all the FOAF in bibliographic records debates.

    I don’t really follow this stuff very closely, and I can’t read the RDA archives, but as an outsider to the cataloging world, FRBR has stood the test of time really well. I can find only one mention of “relational” in the text, and it is only in passing. It doesn’t read to me like something that has that much to do with relational databases. It’s just a pretty clear document that actually tries to understand how to make sense of the data in library catalogs.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s