This whole issue of controlled vocabulary raises at least two
major questions in my mind which the recent discussion of web search
engines only exacerbates.
1. What controlled terms do you use, where do you put them, and how
often do you repeat them?
2. What's the relationship between the controlled language in the MARC
record and the EAD, specifically, and the MARC record and the EADd
inventory as a representation of the collection?
I have much gratitude for the folks who shared their decisions
(and trepidations) in response to my earlier query, and would like to
include my own tentative decisions to what they've said.
Given limited staff, I'm thinking we should include one LCNAF form
of each significant (in terms of late 20th century understanding of
significant) name in a collection -- as UVa. suggested. To each we might
attach a boiler-plate note (perhaps created as an entity or an <extptr>)
explaining that this person appears under various forms of the name, and
instructing the researcher to be adventurous in their searching. I like
the concept of placing the controlled form in context, instead of in a
section of controlled access terms, as our Dynatext software will then
jump to it in the record. I don't think I'm inclined to be rigid about an
"only once" rule, however. Perhaps once per "as-yet-undefined" unit.
But it also seems that names of people and things are easy. It's
when we get to the area of topical subjects (650 terms) that the decisions
get much harder because we don't know where natural language searching
will go, and we can't predict what people will want to research in the
future. At the moment I'm inclined to put some effort into a classical
cataloging model of terms that characterize the entire collection, or
perhaps terms that characterize a series or subseries, but not attempt to
be exhaustive about it. I'm also more inclined to bring out abstract
concepts (politics, practical) over concrete events (elections) on the
grounds that a lot of circumstantial evidence (and perhaps search engines)
will reveal events but may not reveal abstract concepts. But I'm very
open to debate on this.
The other place I'm very open to debate is in the EAD/MARC issue.
If we create a MARC record after we've marked-up an inventory, ought we
transfer our 650 terms from the EAD to the MARC? It's easy to see the
process should go from MARC to EAD, but what about the other direction?
Especially if we're doing controled access terms at both the collection
and series level. Or should we referain from doing that?
If we do catalog at the collection and series levels, and if we
include all those terms in the MARC record, should we expand our 520 and
545 narrative to explain them? Could we instead reference the 856 link to
the URL and encourage the patron to search the inventory for further
infromation on some of the subject headings?
The questions go on and on. Anyone have thoughts? Answers?