On Fri, Apr 8, 2016 at 5:04 PM, Karen Coyle <[log in to unmask]> wrote:

> ....The percentage of things that have decent LCSH assigned to them is
>> small
>> and shrinking for the simple reason that a fewer and fewer humans have to
>> manage more resources.
> I'm not sure what you are saying here -- that there are fewer headings
> being assigned, or that they are not as "good" as ones assigned in the
> past? Or is it that many of our resources aren't covered by library
> cataloging rules?

All of the above. The number of resources continues to grow, fewer people
assign subject headings, and the amount of training those people have
declines. The norm nowadays is for libraries to perform little to no
original cataloging themselves. Especially vendor created record sets are
full of records that lack any LCSH headings, let alone good ones.


> LCSH is relatively flat, the rules for constructing headings are so
>> Byzantine that they stymie even experienced catalogers (which contributes
>> to inconsistent application in terms of quality, level of analysis, and
>> completeness), and its ability to express concepts at all is highly
>> variable as it is designed by a committee on an enumerative basis.
> ?? Sorry, what's this "enumerative basis"?

LCSH is based on literary warrant, meaning that a subject doesn't exist
until needed for an actual item in front of someone at the Library of
Congress or a SACO library. Rather than relate things in a conceptual
universe, LCSH expands on an ad hoc basis. describes the basic
process. LCSH isn't a bad general vocab, but it's not good in specialized
areas because neither the catalogers nor those creating headings have the
expertise to assign helpful subjects. Patrons and staff where I work regard
the headings as unhelpful noise so we don't display LCSH facets (MeSH is a
better fit for our needs).

>   Add to this that concepts in records frequently must be expressed across
>> multiple
>> headings and subheadings, any type of automated assignment is going to
>> result in really "dirty" relationships so I can't blame ILS designers for
>> limiting their use of LCSH primarily to controlled keyword access.
> Well, actually, there's nothing at all "controlled" about keyword access.
> It's pretty much a pot shot, or, as I've called it before, a form of
> dumpster diving for information. There is a huge disconnect between the
> results of keyword searching and the expected functionality (read: user
> service) of controlled left-anchored headings, and I continue to be amazed
> that we've been living with this disconnect for decades without ever coming
> to an agreement that we need a solution.[1] Instead, great effort goes into
> modifying the descriptive cataloging rules, while no major changes have
> been made in subject access. I find this to be... well, stunning, in the
> sense that I'm pretty much stunned that this is the case.

People like the imagery of choosing from a browse list, but they're not
going to guess the left anchored headings because the preferred terminology
and word order often won't be the same as what they're thinking. When they
do type everything right, the absolutely insane numbers of unique
precoordinated subject strings returned would be overwhelming unless there
is little on the subject they seek.

I agree that using subject cataloging rules designed for for filing paper
cards in a computerized environment is insane. But even if the rules were
updated, fixing existing records or creating rich relationships between
LCSH terms would both be impossible. The most practical thing to do is to
do a keyword search on the headings and then return facets based only on
650 $a (i.e. and ignore the rest)   -- which is what most catalogs do.