On Tue, Jan 29, 2013 at 1:36 PM, Ed Summers <[log in to unmask]> wrote:

> If you are thinking of doing name or subject authority control you
> might want to check out OCLC's VIAF AutoSuggest service [1] and FAST
> AutoSuggest [2]. There are also autosuggest searches for the name and
> subject authority files, that are lightly documented in their
> OpenSearch document [3].

This would certainly be a possibility for other projects, but the use case
we're immediately concerned with requires an authority file that's
maintained by our local archives. It contains all kinds of information
about people (degrees, nicknames, etc) as well as terminology which is not
technically kosher but which we know people use.

Ideally, we'd map nonpreferred terms to preferred terms on the fly, but
when you're keyword matching queries that contain more than one
term/concept, it gets a bit stickier. Also, we don't want the vocab to
recommend or map to any term that won't actually lead to an item.

Plus there's always the big catch -- maintenance. Creative stuff is fun,
but reverse engineering clever work later is a real bear. I couldn't agree
with you more that minimizing dependencies and building in elegant
degradation are essential to good design.