GrowthGraphView

MeatballWiki | RecentChanges | Random Page | Indices | Categories

Knowledge grows and proliferates organically. I'm interested in visualizing the process (see MacroScope and http://nike-shoes.cn/ for variations on this theme).

Wikis are great model systems for this project: one page spawns several, these spawn more, and so on. I want to visualize the growth as well as the structure of these knowledge trees, both as an animation and as a static structure.

(Picture a time-lapse movie of the growth of a real tree (an oak say): that's the animation. Now picture the grown tree: that's the static structure. And note--even in the static structure, the thickness of the branches and the topology of the limbs gives us strong indications of the tree's historical processes of growth. That big fork clearly happened decades ago. This cluster of twigs is a recent development. etc.)

Now wikis (and knowledge trees) cross-link in way that oak trees do not (although banyan trees and mangroves do). I want to visualize that too. So I'd like to scale the size of each node by the number of visitors it gets. And I"d like to thicken each edge in proportion to the number of visitors who traverse it. The goal is for major "roads, highways, and destinations" to loom large the way they do on maps and in real life, and for minor arteries and pages to be smaller (and take up less real estate). "Ancestral links" would be distinguished from "foster links"

SunirShah, or CliffordAdams, what would it take to be able to get a data dump that pulls these linkage and usage statistics together? We'd need some historical information about revisions (available, I know) and about a new pages "parent" (i.e., the page from which it was spawned).

The data we need for this would be (roughly) time-stamped event records of...

When a page is born, log its "parent" (the page that introduced the newborn's WikiName).
When an existing page acquires a step-parent (an additional link,) log that event.
When a page is edited, log that event
When a visitor surfs from one page to another, log the identity of the two pages

It would also be nice to have a unique identifier for the human (or ID or IP address) responsible for instigating the event ("rainmakers"?)

In the absence of those historical data, here's something I was able to do from the LinkDatabase

(Detail at [about Wikidot4c.py])

Contributors: JonSchull

To the question specifically addressed at me, what it would take is more time spent coding and less time spent watching hockey games. I've got lots of coding to do, but I don't want to do anything complicated until UseModWiki 1.0 is done. (Isn't it nice when your GodKing can hack the "universe" for you?) -- SunirShah

GrowthGraphView

Discussion