On 5/15/07, Ted P Gemberling <tgemberl_at_uab.edu> wrote:
> Jon,
> There was some discussion of that approach on the PCC list last year.
> The problem is that dates are much more likely to be unique than
> subjects. Especially since subjects are somewhat general. If there's a
> "Smith, John, writer on astronomy" set up, it won't be long before
> there's another person who'd have to be described the same way. Of
> course even dates aren't always completely unique. Occasionally you have
> to add the exact date. For example, there are three John Smith, 1924-'s,
> so two of them have the month and date added at the end.
I think you misunderstand me. I'm not proposing replacing techniques
for determining identifiers, I'm proposing addition of known
information to the display. Since we can know "Smith, John 1943-"
wrote, since he's now a bibliographic identity, we can automate some
extraction of information from the bibliographic records that use that
authority record. Yes, it's not always going to be perfect, but a
combination of various author information may help a person quickly
distinguish in many cases.
>
> Your proposal is really already in effect in "undifferentiated" name
> headings. If you can't get any more information about several people
> named John Smith, you put them all on one authority with descriptions of
> the different things they wrote. The undifferentiated heading for Smith,
> John is a really good example of that. When more information is found
> about one of them, that "bibliographic identity" is removed from the
> undifferentiated authority.
Well, no, since what I'm proposing won't work very well with that. It
would list a seemingly random clump of titles. I'm saying when we
identify a bibliographic entity, we then show users some of the
information we have about that particular bibliographic identity.
Namely, the information in records with that identity in it as a
principle author. We can combine this with sampled circulation
statistics to get a better picture.
My approach isn't doing anything with the underlying data, just
utilizing the connections that are already there. An undifferentiated
heading, mistakes in one, or an uncontrolled one
aren't likely to be assisted by this technique, but distinguishing
between two identified bibliographic identities with the same name but
just date or is made easier without the user having to "drill down".
Worse case, the user has to start clicking through all the various
"Smith, John 194x" who have astronomy in a lot of their bib records,
but that might eliminate several other Smith, John 194x".
We're able to do much more with techniques like this with our current
generation of processing power and software techniques, and it seems a
pity to let it go to waste.
Jon Gorman
Received on Tue May 15 2007 - 14:49:06 EDT