> On Tue, 13 Jun 2006, Thomas Arendall-Salvetti wrote:
> Agreed. Let the searcher put a whole bunch of keywords in a box and have a
> smart system that takes that search and uses real authority control,
> relevancy ranking, etc. to bring back not only what user may think is
> "good enough" but also what's truly helpful. Google does this better than
> Amazon but I can't say either does it well. They do have the content,
> though, and the big name familiarity.
I think one important idea (I'm sure this has an official name I don't
know) is returning the results with a profile/summary of the results, that
can be used to fine-tune and select from within the results. Do you want
things on topic A, or on topic B? Do you want things in audio, or video?
Some of those keywords matched an authors name (and/or a particular author
is well represented in the search results for other reasons), do you want
everything by that author (or only things by that author?).
Google is adding a little bit of this with Google coop, if you choose to
'subscribe'. Other search interfaces try to do this with
machine-generated 'clustering'. With our rich human-applied metadata, we
should be able to do this really really well. The NCSU/Endeca interface
tries to do this, with in my opinion mixed success (although of course
it's a great first step). Part of the problem is that, while one can say
our rich metadata _should_ allow us to do this really well, in fact our
metadata is not neccesarily suited for it. (My favorite example is how the
NCSU interface gives you a 'genre' facet, in which 'fiction' is one value,
but it doesn't neccesarily in fact correspond to all fiction in the result
set). Another part of the problem is actually technical, figuring out how
to provide a powerful interface for this that is not overwhelming or
confusing.
--Jonathan
>
> Mary
>
> Mary Grenci
> University of Oregon Libraries
> mgrenci_at_uoregon.edu
>
Received on Tue Jun 13 2006 - 15:57:32 EDT