... synthesis.[*]
An index page is a page consisting of links to a set of pages that cover a particular topic (e.g., electric guitars).
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...[5][*]
http://www.cs.cmu.edu/webwatcher/
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...[7][*]
http://zeus.gmd.de/projects/avanti.html
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... Firefly[*]
http://www.firefly.com
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... site,[*]
http://machines.hyperreal.org
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... server.[*]
A web site is restricted to a collection of HTML documents residing at a single server -- we are not yet able to handle dynamically-generated pages or multiple servers.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... visitor.[*]
In fact, this is not necessarily the case. Many Internet service providers channel their users' HTTP requests through a small number of gateway machines, and two users might simultaneously visit the site from the same machine. Fortunately, such coincidences are too uncommon to affect the data significantly; if necessary, however, more accurate logs can be generated using cookies or visitor-tracking software such as WebThreads.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... visitor,[*]
We consider an entire day's page views to be one visit, even if a user made, for example, one morning visit and one evening visit. This simplification does not greatly affect the data; if necessary, however, the series of page views could be divided at significant time gaps.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... cliques.[*]
Note that we place a maximum size on discovered clusters not only in the interest of performance but because large clusters are not useful output -- we cannot, practically speaking, create a new web page containing hundreds of links.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... site,[*]
http://machines.hyperreal.org
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... period.[*]
Data sets are publicly available from the authors.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Mike Perkowitz
1999-03-02