On 5/24/07, K.G. Schneider <kgs_at_bluehighways.com> wrote:
> It doesn't have to be either-or. Computer-assisted classification could
> provide huge bang for the buck in terms of the ability to tackle the amount
> of content out there while providing good quality metadata; let the
> computers make the first pass, then clean up their problems and add some
> human sensemaking.
This is the tact PINES takes with Evergreen meta-records (FRBR-ish
groupings)--machine algorithm to group records, but human touch to
tweak them where needed.
We also do this for deduping; a machine algorithm gets us close, and
then we farm out the proposed groupings to catalogers for approval
(they get to see the MARC side by side and can pick the lead record to
merge to).
> That would be a good balance. That keeps us in charge
> (machines should always work for us, not vice versa... I learned that from
> Asimov) but gives the computers the grunt work, which frees us to do more of
> the intellectual work.
Like figure out psychohistory. :D
--
Jason Etheridge
GPLS -- PINES Development
http://open-ils.org/
Received on Thu May 24 2007 - 18:54:41 EDT