snowcrash network.

the data was collected from amazon.com by hand in the form of a tabbed outline. the outline is 1110 entries long.

i then wrote a python script to turn this outline into XML so it could be read into prof. Akiran's Processing applet. here are the first few lines of the outline, next to the xml code. it's all pretty sickeningly redundant.


while i was finishing up the data collection, i realized the outline could be way scaled down using a lindenmayer system, where you have rules for each term that is not at the deepest slice. the rules in this case would just consist of appending the term with an indented substring of its "immediate children." this way, a complex network such as the one above can be represented by something much more compact and much easier to comprehend. for example, in the outline format, itself more transparent than XML, one still cannot directly see the layers of self-similarity in the network as the result of closed loops. i was made painfully aware of the resulting redundancy when i assembled the outline, mostly via cut and paste. with an l-system one could simply define each node's children, the axiom, and the level of magnification.