Sunday, March 27, 2005

Random thoughts on XOBIS

Kevin Clarke, one of the authors of XOBIS, kindly left a comment on my recent blog post on the topic. It shamed me into returning to the XOBIS general overview document I peeked at briefly when originally writing about it. I've now given the entire document a quick read. I can't claim to have an in-depth understanding of it at this point; it certainly took me several readings of the FRBR report and a decent amount of time thinking about modeling different things in FRBR before I felt I could really say anything intelligent about it. Nevertheless, I have a few initial impressions on XOBIS.

The most obvious difference I see between XOBIS and FRBR is that XOBIS attempts to be a model that can describe all of knowledge, while FRBR limits itself to modelling bibliographic relationships. In a practical sense, for recording bibliographic data (and this certainly isn't the only possible use of XOBIS!) this means that XOBIS explicitly handles entities that represent in a bibliographic environment creators or subjects of bibliographic items (and, in FRBR, other Group 1 entities), currently residing in a relatively unstructured way in name and subject authority files. FRBR, on the other hand considers only briefly its Group 2 ("person" and "corporate body") and Group 3 ("concept," "object," "event," and "place") entities, focusing instead on Group 1 entities.

Relationships between entities is a key feature of XOBIS; they are also a bit confusing to me on my first read. My initial impression is that the relationships as specified focus more on subject-type relationships rather than relationships among bibliographic items. My reading is that the XOBIS definition of work is much closer to what we currently consider a bibliographic item than FRBR's work. The discussion and examples in the overview document talk about versions of works and how they are related, but I saw much less about the "accidental" sort of relationship a FRBR-ish work (as its expressed in a specific manifestation) would have to another expressed work on the same manifestation, for example, two symphonies appearing on the same CD. It would be an interesting excercise to map out how the XOBIS model would handle this sort of situation, where the symphony itself is the entity of primary interest to a majority of end-users rather than the specific performance or the title of the CD on which it appears.

XOBIS comes out of the Medlane project of the Lane Medical Library at Stanford. I wonder what effect medical materials have had on the development of the XOBIS model. I know my focus on musical materials in various projects, most notably Variations2, certainly strongly affects my thinking about FRBR and related efforts. I'm sure that's obvious from my earlier question wondering how XOBIS would handle a situation that the Variations2 model is designed around.

There are also some very interesting items in the report's bibliography, including a project mailing list (renamed since the version listed here, and looks low-traffic). Time for citation chasing!

3 comments:

dick@stanford.edu said...

Jenn, I appreciate your comments on XOBIS. Although you raise many issues, I couldn't resist a comment or two.

"... the relationships as specified focus more on subject-type relationships rather than relationships among bibliographic items."
XOBIS treats all relationships equally, leaving emphasis up to the cataloger. It actually has unlimited possibilities for bibliographic relationships, such as "Based on", "Parody of", "Cited by," as well as the typical ones. On the other hand, it embraces other relationships, such as those between two authors ("Colleague" or "Husband"), between an author and a concept ("Discoverer"), between two places ("Capital", between a word and a person ("Coined by"), etc., etc. All of these relationships can potentially be under authority control and in turn have their own relationships. It does make a rather neat little package considering the scope.

As to the item vs. work, in XOBIS both title authorities and bibliographic titles live in the same structure. A 'uniform title' authority behaves like an umbrella record. There is an example of Dame de Pique that explores the authority/work/version/performance issue that may help with your musical interests. (Without specifics this is a very large topic to tackle.)

We deliberately generalized XOBIS to attempt deal with fundamentals and especially to help address the museum/library continuum. Of course, it probably shows that this effort was made in a medical library. :-)

Our principal bibliography is http://medlane.info/bibliography.php and for a more accessible intro, try CCQ article http://elane.stanford.edu/laneauth/XOBIS_CCQ/XOBIS_CCQ.html

I deeply appreciate that your scratching more than the surface.

Regards, Dick
Another of the authors.

Jenn Riley said...

Thanks so much for the comments, Dick! Sorry it's taken so long for me to respond. The usual apologies about travel, deadlines, etc., apply.

I think my statment "... the relationships as specified focus more on subject-type relationships rather than relationships among bibliographic items" you quote probably should have been written, "the subject-type relationships in XOBIS are more obvious to me on first reading than relationships among bibliographic items." Coming at the XOBIS initiative from a FRBR-ish perspective, I first looked for how those sorts of connections would be handled, and got a bit bogged down in the other stuff. But that's my approach to the first reading, certainly not a good reason to structure a report a certain way! :-) I'm sure the potential of bibliographic relationships would emerge more clearly to me on subsequent readings.

I need to look at the uniform title authority in XOBIS more carefully. UTs are great for a lot of things, but we've found in Variations2 they don't always make obvious the relationships we'd like them to. Something to think about, for me.

Anybody from XOBIS going to be at the OCLC FRBR meeting on May 2-4?

dic said...

Jenn, If you would send me some examples of the specific relationships that aren't being handled adequately by UTs, I would enjoy teasing them about from a Xobian perspective.