Tim, in part I think at one point you confuse LCSH and LC
Classification. LC Classification shelves things in a single place; LCSH
allows multiple subject headings to be added to a record.
Apart from that, if you are looking at ranking, a few bits of info:
1. When creating a MARC record, the first LC subject heading on the
record is supposed to be one-to-one with the single LC classification
number assigned to the item. I don't know if catalogers still do that,
but it was true at one time. That would presumably make the first LCSH
field be "more important" than the others.
2. The LC headings are both broad brush strokes and overly subdivided
statements. Broad because they only address the overall topical thrust
of the item [Medicine -- social aspects, on a book about how medical
science has treated men and women differently as patients]; overly
subdivided because LC (so they once claimed) subdivides topics such that
no one topic gets more than 200 items under it. This means, of course,
that some topics are much more subdivided than others, cf. anything
under "united states - history - civil war"
[United States -- History -- Civil War,
1851-1865 -- Hospitals -- Periodicals; United States -- History -- Civil
War, 1861-1865 -- African Americans -- Juvenile fiction.] as opposed to
[Body piercing -- Pictorial works]. However, it seems that today there
is a lot of subdividing going on, and in a browse in a large catalog a
huge number of headings have only 1 or 2 items, including "Punk rock
music -- California -- Berkeley -- History and criticism" (with one item
in MELVYL). _ _ I did hear that at around the same time that the MARC
record became part of automated systems (and no one was trying to
squeeze subject headings into the little margin at the top of cards)
that the average number of subdivisions per heading AND the average
number of headings per record went up measurably. So older works would
have less to work with.
I'd like to see some playing around with the LC class numbers. I think
this would be difficult, but the classification is in machine-readable
form and there is text associated with the numbering system. The
advantage is that you have a real hierarchy, or I should say "some real
hierarchies" because there isn't really an overarching one. You also
have facets for things like geography and time.
kc
<http://melvyl.cdlib.org:80/F/JRP876T1T711L3726TX2E65JSVJTU8Q3CHVKXE1EVBF8YXYYLM-03150?func=find-acc&acc_sequence=035021245>p.s.
But what really gets me about the LC headings on records is exemplified
by this: The 1948 edition of Norbert Wiener's book "Cybernetics" has
only one subject heading: "Mathematical statistics." Since he had just
invented the term "cybernetics" it didn't exist as a term. By 1965, when
the book was revised and reprinted, the only subject heading is:
Cybernetics. Nothing links the two, at least not in the subject heading
world. In this example, user tagging would probably be more useful.
(Note, the books do get similar shelf numbers in some of the libraries I
can see on my screen.)
Tim Spalding wrote:
> I just wrote up a blog post about trying to tease relevancy ranking
> from LCSHs:
>
> http://www.librarything.com/thingology/2007/02/can-subjects-be-relevancy-ranked.php
>
>
> I wonder if anyone has made, seen or can think of any good methods to
> do it. So far I've only seen non-ranked and popularity-ranked results.
> In the blog post I talk about playing around with how LCSHs reinforce"
> each other statistically, but I couldn't get the algorithm to produce
> good results more than sporadically.
>
> I'm not sure if this is a cataloging or a coding. Maybe that's the point.
>
> Tim
>
>
--
-----------------------------------
Karen Coyle / Digital Library Consultant
kcoyle@kcoyle.net http://www.kcoyle.net
ph.: 510-540-7596
fx.: 510-848-3913
mo.: 510-435-8234
------------------------------------
Received on Mon Feb 05 2007 - 16:31:08 EST