Casey Bisson wrote:
> But your real point is that we need find a way to group results into
> more generalized categories. That's where I admit my limits as a
> programmer and step back.
>
I think the path to this is in the BT (broader term)/NT (narrower term)
"cross-reference" relationships which _are_ included in (much, but I
don't think yet all) LCSH, as of a couple decades ago. Although they are
not usually shown to the user at all.
With Project Fred, you lookup the authority records to access those BT/NT.
Of course, while they are in LCSH, the fact that they've never been used
for much of anything useful in a mass-processed way (just cataloger
guidance), means that who knows if they will really work. I think many
of them were in fact machine generated, although they are human
maintained. Theoretically.
When we start looking at LCSH, and seeing _how many_ ways relationships
between headings are indicated, we are reminded that LCSH was designed
for a very specific environment, where all headings would always be
presented only in an alphabetical list, and gives us quite a bit of
trouble in our present environment where we want to be more flexible.
Some of the ways LCSH has for indicating relationships between
terms/headings include (but probably arne't limited to):
1) Sub-division hiearchy (fixed and free-floating).
2) NT/BT
3) "Inverted" headings, such as "Cookery, Indic", which is a subset of
"Cookery" (one relationship), but also related to other ", Indic" headings.
4) Direct order headings which are intended to be grouped alphabetically
with related headings, but have no other indicator of relationship other
than common left-anchored string.
5) Parenthetical qualifiers, geographic and other.
There are others. I think someone, I forget who, has an article
analyzing 'syndetics' in LCSH, which covers some of these (but doesn't
cover others I thought of too).
Some of these are quite good at displaying relationships to a human
browsing an alphabetically arranged list---but are nearly impossible for
software to mechanistically extract meaning from, so that meaning can be
used to provide different interfaces.
Jonathan
> Note about the software: It's Scriblio (formerly WPopac). More info
> at the links below.
>
> http://maisonbisson.com/blog/post/11133/
> http://www.techsource.ala.org/blog/2006/12/unsucking-the-opac-one-
> mans-noble-efforts.html
> http://www.techsource.ala.org/blog/2006/01/library-20-in-the-real-
> world.html
>
> --Casey
>
>
> On May 9, 2007, at 1:23 PM, Ted P Gemberling wrote:
>
>> I think traditionally,
>> "faceted subjects" meant subjects that have been "decomposed" into
>> separate logical elements that are extremely general, so that it's
>> easy
>> to move from one combination of the elements to another. I suppose
>> maybe
>> the "decomposition" is present on Lamson's Endeca-type page with the
>> distinction between subjkey, author, format, and Meta. But the
>> subjects
>> themselves in subjkey are not faceted in that traditional sense.
>
>
> On May 8, 2007, at 5:30 PM, Kristin Antelman wrote:
>
>> Your example, sociology of education, demonstrates the problem. In
>> our
>> catalog, a keyword search on that term gives 861 hits, and the facet
>> linking to the correct term, educational sociology, has 388 hits. If
>> you look in our LCSH browse index, however, you will find 618 items
>> with
>> the heading "educational sociology" and its associated 81 subheadings.
>> So the keyword searcher is not seeing 230 items with the exact heading
>> they were searching for.
>
>
>
> Casey Bisson
> __________________________________________
>
> Information Architect
> Plymouth State University
> Plymouth, New Hampshire
> http://maisonbisson.com/blog/
> ph: 603-535-2256
>
--
Jonathan Rochkind
Sr. Programmer/Analyst
The Sheridan Libraries
Johns Hopkins University
410.516.8886
rochkind (at) jhu.edu
Received on Wed May 09 2007 - 12:21:52 EDT