audioscrobbler experiments

1. bands like the bands i like

was listening to nofx this afternoon and was wondering what similiar bands exist that i've never heard of?

audioscrobbler, the back end for last.fm, has exposed a vast array of recommendation data as a webservice.

time to write some code to make use of it....

wrote a ruby script fetch.rb to mine the data (what did we do before hpricot?)

run it as ./fetch.rb nofx 75 20 > nofx.yaml

• will start at nofx
• extract artists that have a similiarity higher than 75 (on a scale of 1 to 100)
• for thoses artists, download their most similiar artists
• continue until 20 have been downloaded in total

this builds a yaml file of a graph where nodes represent artists and edges represent a similiar artist.

now to draw it! another script generate_dot_input.rb to do some collating of data and generate a .dot file for graph viz to use.

run it as ./generate_dot_input.rb < nofx.yaml | neato -Tpng > nofx.png

here are two graphs centered around nofx generated at different depths/widths;
a one small (150k) generated with 75 min similiarity and 20 nodes max and
a one big (1,300k) generated with 60 min similiarity and 45 nodes max