Visualization of Blogspace

Jon Schull's Weblog

JS Weblog Home

recent bio

JS on one page

email

Visualization of Blogspace

Visualizations of Blogspace.

Stepan from Blogtree kindly provided me with a bunch o' data from his BlogTree.com, blog genealogy site. Each registered blog owner records the pre-existing blogs that inspired their own blog. I used python to crunch the data and re-cast for graphviz. Graphviz lays the graphs out and renders them. These images are larger than they appear on this web page. View them in a new window for more detail.

	On these graphs, lines radiate from inspirer to "inspiree", with an arbitrary blog chosen as the very center.
	Here's a subset of the data. Now it turns out that one blog looms large in these data, as it does in Blog space. This is the blog colored red in the lower right corner of the big round component. (You'll see it when you zoom in)
	In the third figure, I put that blog--Dave Winer's Scripting News--in the center, and scaled the fonts so that the area subtended by a character is in logarithmic proportion to the number of descendants

Now we transition from 'art' to' science.'

While the images above are very evocative of one of the great phenomena (the proliferation of memes) but they don't really allow you to read the history. Rather, they make it clear that significant distillation is going to be needed to make these pictures tell the story.

I've done some of that, and this is shown in the next few images.

The distillation techiques used are (1) color (2) scaling of line width (3) scaling of font sizes (4) conflation. (Conflation is my name for a process of putting all childless blogs of identical parentage in a single clump. This greatly reduces the amount of ink needed to depict the relationships, leaving more unused space for still more nodes. The technique could be extended to connection patterns other than "childless siblings of the same parents"; that was simply the easiest and most frequent.)

	This demonstrates color and scaling. Incidentally, these diagrams represent a particular "focal node's" progenitors and offspring. The focal node is in red.
	This example shows conflation. The triangular "piles" of "sibling orphans" are clumped to occupy a single position.
	This uses all the techniques, and shows one of the more prolific trees. Combined with a fast, responsive, cacheing, zooming image navigator...this could really become something useful. (I've tried SVG browsers but they just don't seem responsive enough). There is much more to do of course. In particular, these graphviz-generated layouts are optimized to minimize line-crossings. Resulting in the vertical "thorougfares" through which so many lines pass. Unfortunately this disrupts the continuity of the lines. Line crossings would be more tolerable when lines are colored as here and tapered (as I'd like to see). The botanical branching patterns of botanical trees are what we want, I think. We're working on it.