An ideal WikiPortal is a page that links to content of interest to the visitor. While the name has a bad connotation because commercial web portals pollute the information offered with advertisements, WikiPortal""s denotate pages that are natural portals.

Natural portals can be found by finding pages that link to a lot of other pages perhaps by an AccumulatedRandomWalk or a DepthSearch with depth of maybe 3. The usefulness of a portal decreases if it refers to too many pages, so AllPages does not qualify as a WikiPortal.

Ideally, a list of WikiPortal""s would be generated that allow to reach the entire Wiki:

Start with a list of AllPages and an empty list of WikiPortal""s.

Among the pages in the list, find a NaturalPortal as page that allows you to reach the most pages in the Wiki in a few steps, but skip those pages that have IndexNature (defined as reaching more than 25% of the Wiki, maybe ). Add the NaturalPortal to the list of WikiPortal""s, and remove all nodes that are reachable by this portal from the lists of pages in a few steps.

Repeat until the number of pages reached by the "NaturalPortal" in a few steps falls below a threshold. Maybe create an artificial portal for the pages remaining in the list.

--PeterSchaefer

I could change ShortestPathPages to be iterative.

   while there are pages left
        centre = mostCentralPage()
        add centre to portal list
        remove all pages within N steps of centre
   end while

   emit portal list

where <tt>mostCentralPage()</tt> is the number one entry on ShortestPathPages. -- SunirShah

CategoryIndexingScheme