[In response to my question: "Are you saying that if people search for John Huston as a film director in our catalogs, they should not expect to find the films in which he was a director?"]
On 24/09/2012 15:02, Amanda Xu wrote:
There are four kinds of cases that we need to act upon so as to add a relator code to 100, etc. fields in a bib consistently across result set including film director:
- on-fly rendering of relator code for existing MARC data being asked for use based on rules that many discussed here already and MARC leader 06, 07, etc.
- newly-created RDA data will have such code stored in the database and ready to be used, and met the objectives setup for FRBR and others that we’ve discussed;
- recon highly circulative or frequently used MARC data so that it is compatible with RDA data, including relator code, and improve performance overtime;
- program changes for newly created RDA/FRBR data to be reused, interchanged, discovered and others among upstream and downstream data utilities that would consume or publish the data, including RDFa conversion for Web search engines;
If we can, e.g. time and resources permitting, we can run four cases at the same time collaboratively. If not, let’s focus on case 1, 2, and 4.
That’s how I see how we can make film director and other relator codes to be implemented for RDA data consistently if our existing data have designated materials for motion pictures, sound recordings, etc. consistently.
A lot of value-added processes have been added to library metadata, e.g. MODS, MADS, etc., which might be leveraged on-fly RDA generation. But you know the danger when you are not cited from the original.
We also want existing and future workforce to be ready for the transition. Catalogers have to be concerned and work in parallel modes, e.g. MARC and RDA/FRBR.
Concentrating on 1, 2, and 4 ignores the problem of the so-called “legacy data” (also known as 99.999% of everything we have).
So, I guess your contention is that changing older records to add the role “film director” can be done automatically, although your point 3 mentions recon (manual) of selected records. Assuming that automatic updates can work seems to be making a highly unwarranted assumption and should be subject to at least some kind of preliminary project to see if what it produces is not completely incoherent, which is in fact, what I believe. In this respect, I remember when I was working with optical character recognition and found that it was rated as somewhere around 95%+ accurate, but when you see the results of OCR, you begin to question what that 95%+ accuracy means since the final problem is often total gobbledygook.
Additionally, this is only for a single role. There are actors, writers, producers and other kinds of roles available. Also, this is only a single format and in a single language.
Some may say that this is “being negative” but it seems to me to be realistic and to ignore it is “being idealistic”. I wish it would all work out too, but from the very first day that a search is implemented for specific roles, we know what the public will see. This is nothing to do with “maybe” or “once in awhile”, but it has to do with “definitely” and “in practically every search”. Shouldn’t there be at least something to say to the public who will begin to get strange search results?
Or do we just leave all of that to public services, who will have to deal with the fallout?