There is a project underway at the University of Maryland at College
Park called CLiMB (Computational Linguistics in Metadata Building).
Info about it is here:
http://www.umiacs.umd.edu/~climb/
This project is designed by academics in the fields of computational
linguistics, library science, and computer science. They were looking
for sets of images with text to use to test the software they developed.
We gave them access to the electronic version of the catalog of fine art
in the Senate--images, with descriptive text. They ran the text through
their program, which is linked to 3 of the Getty Vocabularies (AAT,
ULAN, TGN), and the program automatically supplied subject terms and
names of artists and geographic headings. The curatorial staff here at
the Senate were absolutely delighted--the program seamlessly suggested
terms from controlled vocabularies that are used in their field. The
terms can be accepted, or not, so humans always are in control.
I think this is very cool stuff. It is not a commercial application and
I don't know if plans are in the offing to make it commercial, but I can
definitely see the utility of it. If they could get licenses to other
controlled vocabularies to load into the program, it could really make a
huge difference in cataloging electronic text. For instance, the Senate
materials I catalog are available electronically. If the CLiMB program
was linked to LCSH, and the program was run against the text of
hearings, the program could suggest more, and quite possibly better,
subject headings than I could come up with on my own from physically
eyeballing the hearing, the way I do now.
I'm not saying such programs could or should supplant people, only that
they can definitely augment what people can do.
Betsy Moon
Cataloging Supervisor
U.S. Senate Library
SRB-15, Senate Russell Office Bldg.
Washington, DC 20510
betsy_moon_at_sec.senate.gov
202-224-5581 (phone)
202-224-0879 (fax)
Received on Fri Aug 31 2007 - 09:38:03 EDT