[Sigia-l] "Best Bets" or "Accidental Thesaurus" Examples?

[Avi Rappoport]

I think manual recommendations (Best Bets is a Microsoft term) for 
common searches are a great idea and am suggesting them to all my 
clients.  There may well be false matches, but that's common in 
search engines in general.  Last time I went to the MS site and 
searched for "Bill Gates", the #1 search engine hit was from 1997, 
technically the "most relevant" according to their algorithm.

This is not to say that we shouldn't improve search relevance -- of 
course we should.  But that's really hard, given the preponderance of 
short queries and the Zipf distribution of queries (many common 
queries, many unique queries, not much in the middle).   Using human 
judgement to supplement algorithmic relevance for the common queries 
is simply using the appropriate process for the problem.  This is not 
a case for theoretical purity, it's a case for good heuristics.  I'd 
love to get results on a usability study for people who type in 
common queries like "bookstore" and either do or don't get 
recommendations.

As for examples, Dell and HP both have implemented Recommendations. 
I know that the tools are built into both Inktomi Search Software and 
the Google Search Appliance, but I don't remember offhand which 
customers have used it.

Avi


At 1:49 AM -0700 8/26/02, Richard Wiggins wrote:
>My simple suggestion is that, for the most commonly sought content, an
>editor in East Lansing can do a better job of helping local users than can a
>robot in East Lansing or in Palo Alto.  No search engine is perfect. Neither
>are editors perfect, but when consistently hundreds of people at a
>university search site enter "bookstore" it doesn't take Harold Ross or Miss
>Cleo to divine what the reader seeks.

-- 
Search Server Industry Analysis from Search Tools Consulting
    (510) 845-2551  -- <mailto: analyst at searchtools.com>
Complete Guide to Search Engines for Web Sites and Intranets
    <http://www.searchtools.com>