Here is the algorithm for Analysis of Wiki Link Structure:

- Download from Ward's server the file containing all links.
- Ignore all links to external sites, and all "links" to non-existent pages. For the purposes of finding potentially valuable but hard to access pages, these "links" are irrelevant.
- Consider removing from consideration all pages with CategoryHomePage. Some pages only have signatures as their out-going links. Removing all home pages should make more pages into dead ends, and more pages into orphans, but will no longer list home pages in the output.

- Remove from the graph all vertices listed on CategoryAutoIgnore. These are pages that should probably be ignored when trying to understand the linkage structure of the wiki. There is some discussion about this issue on that page.

- In the remaining graph, find all vertices with zero out-degree. List them, then remove them from the graph. Repeat this process until all vertices have out-degree > 0.

- In the remaining graph, find all vertices with zero in-degree. List them, then remove them from the graph. Repeat this process until all vertices have in-degree > 0.
- The graph that now remains is such that all vertices have non-zero in-degree and non-zero out-degree.

- Identify all small (up to 100 vertices) sets of vertices such that the set is strongly connected and has no incoming links.

- Identify all small (up to 100 vertices) sets of vertices such that the set is strongly connected and has no outgoing links.

EditText of this page (last edited January 2, 2011) or FindPage with title or text search