[Sigia-l] "best bets": trademarked? other names?

[Avi Rappoport]

At 4:05 AM -0400 9/30/02, PeterV wrote:
>Why do we want to indicate the way these specific results were achieved?
>
>I think in most cases (assuming a non-search-expert user), integrating
>best bets with the rest of the search results (why not just put them on
>top?) is the best way to go. The interface problem is part of this: many
>"best bets" implementations present the results separately, and I have
>the feeling these results often get skipped because they don't *look*
>like search results. I know I do that.
>PeterV

I'm reluctant to do this because of the value of what Jarod calls 
"the scent of information".  Google and other search engines use 
context "snippets" which make it easy to understand the why of a 
search result, and I think that's incredibly useful, saving tons of 
time and clicks.  Moving the recommended results into the general 
search results makes it much less clear what's going on.

That said, I'm a huge fan of allowing enterprise search engine admins 
to weight certain types and locations of documents so they come up 
before others.  Term frequency is a slender reed to rely on, and 
adding heuristics (such as "title is important") are only of limited 
help -- you get a lot of virtual ties in relevance scoring.  In one 
project I did, many of the first 10 results for searches I tested 
were from the online conferencing area, *not* good results (but short 
so they did well on TF).  So I recommended that the company add 
weight to URLs for product pages, FAQ answers, and other valuable 
pages.  It didn't break the scoring system, it didn't throw out the 
conferencing (which in some cases was useful), it just used human 
judgement as to areas which were more likely to be helpful.

Avi

-- 
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>