Wed, 10 Feb 1999 09:17:10 -0800
text/plain; charset="iso-8859-1" ; format="flowed"
At 10:28 PM +0100 2/9/99, Heike Röttgers wrote:
> I am actually analyzing search engine logfiles in order to get some
> information about user
> information retrieval behaviour.
> I would like to know if anyone has some experience or tips for this
> kind of research.
One approach to concentrate on is the searches themselves. Look at
how many words were used, whether the searchers used Booleans or
other commands, whether they took advantage of the features in the
It's a good idea for you to re-issue the searches and see what the
results look like -- the number of matches is very interesting, but
also the kinds of matches. How relevant were the results, on the
first page in particular? Are the pages you consider best at the
top? Can you see any regular patterns on better matches vs. worse
matches that would improve the relevance weighting algorithm?
The other interesting data comes from tracking user behavior, which
is a lot harder. Follow the users by IP address or host name and see
how many results they look at. That's a very interesting number!
Looking at the pages they chose in conjunction with the query may
tell you something about how these particular searchers read the
titles and description/extract text and made their relevance
judgements. On the other hand, you may learn that most users just
try the first four results... depends on the system. I also like to
look at the next searches from the same address, and guess if they
are trying a new search on the same topic with a different approach,
or if they are going on to another topic. This takes time and
judgement, but is worth trying.
Hope this helps -- please keep me informed about your progress!
Web Site Search Industry Analysis from Search Tools Consulting
(510) 845-2551 -- <mailto: [log in to unmask]>
Guide to Web Site Indexing and Local Search Engines: