The Accidental Thesaurus

[Richard Wiggins]

Folks,

There was a thread a couple of months ago about using the actual search
terms typed by users to augment a thesaurus of library terms.  I'm looking
for examples of libraries, universities, and others doing similar things.  

We've been doing something even more heuristic at Michigan State.  We have
built an MSU Keywords service that matches user searches with the best URL
we can find on a given topic.  

We started with a small list of official sites and surmised the best
keywords for each.  From there, this is now almost entirely user-driven.  We
examine search logs to see what terms people are searching for the most, and
we make sure every popular search input is in the database.

When a user enters a search at search.msu.edu, the MSU Keywords database is
searched.  If there are matches, hits appear before AltaVista results.  If
there are no matches we just show AltaVista hits.

An example of our search logs is at:

http://search.msu.edu/info/logger.html?when=last31days

A daily link checker also informs me of any broken links in the database:

http://keywords.msu.edu/info/showbadurls.asp

We accept misspellings and strange variants of official terms.  You type
"libary" and we'll get you to the "library" (though the right spelling will
appear on screen).

Have any of you created a similar critter for your library?  Either as a
front end to your library Web site search or a front end to something else?

Another university site doing something very similar is Berkeley:

http://www.berkeley.edu/

Just do a search off the home page for "human resources".  They don't even
present Google results if they find matches in their official registry.

Following are other examples of this done by businesses.  There must be some
more library and university examples.

-- AT&T.  In the past if you searched for "long distance" you got a press
release, not a page for rate plans and signup.

-- ESPN has a very similar concept.  Go to espn.com and type "Vitale" to see
how it works.

-- Probably the most famous application of this idea is AOL Keywords.  A
single search box serves to match users to registered AOL Keywords, or to
broader searches, or to Web addresses.   

Know of any other good examples?

/rich

PS -- I was trying to think of a good term to describe my seat-of-the-pants
approach, and came up with "accidental thesaurus."  I did a Google search to
see who else might've used the term, and I found an example: me!

http://library.usask.ca/access98/ppoint/wiggins/html/sld079.htm

_________________________________________________________________


PPS: Quote from earlier Web4lib thread:

...

I am looking for other examples of this approach.  

___________________

It's not a question of "first" -- not many people are doing it and 
there's no research on improved search experience that Lydia or I 
could locate! So there is plenty of scope for you :-)

Avi

____________________________________________________
Richard Wiggins
Writing, Speaking, and Consulting on Internet Topics
rich at richardwiggins.com       www.richardwiggins.com