[Sigia-l] Re: Faceted Classification

[Laura Norvig]

Thank you, Gene, I think you've isolated a key point here, which is 
that faceted search may work best when users are prompted to choose 
terms from a controlled vocabulary for each facet.

Also, as I think you're saying, it is not really faceted 
classification unless the vocabulary IS controlled. At least, that is 
my understanding, although it seems a combined approach, where some 
fields, e.g. publication type, could only be searched with controlled 
terms, but other fields, such as Title, could be searched with free 
text, might work.

And to the person who asked for some sample faceted searches, this 
has been mentioned several times but do check out Marti Hearst's 
Flamenco project at
http://bailando.sims.berkeley.edu/flamenco-interface.html

Laura Norvig, MLIS
"Intelligent Human Agent"
lauran at etr.org


At 10:35 AM -0600 7/11/02, Gene Smith wrote:
>  > From: "Christopher Fahey [askrom]" <askROM at graphpaper.com>
>>  It has always seemed to me that "faceted classification" is just a fancy
>>  way of saying "put your content into a database and use lots of metadata
>>  fields".
>
>I don't think that's quite right.  My mental model of a *simple* faceted
>system is a database table where the metadata fields for each record are
>actually look-up fields from which you can pick just one term (rather than,
>say, open metadata fields where you could enter anything you want).  For
>practical purposes, some metadata fields should be left open, such as
>date/time fields, or fields that contain scalar values like price or number
>of pages in a book.  If I'm wrong about this, though, I'd really appreciate
>a correction.
>
>I've always had the intuition that faceted classification works really well
>for homogeneous and well-described domains (e.g. wine), but wouldn't work so
>well for domains with broad and diverse content (e.g. think of developing a
>faceted classification system for the programs and services of the US
>government).  In that last example, the LIS folks might say it's possible
>(maybe it's already been done), but an IA/UX person would probably ask first
>if it's useful for their target users.  I think that's a key difference
>between the information-centered approach that Bates outlines and a
>user-centered approach.  But that's a quibble more than anything
>else--overall I thought her article was excellent.
>
>Cheers,
>
>Gene
>
>
>
>-------------------------------
>gene at atomiq.org
>http://www.atomiq.org