25 August 2006

Radial-ontologies math quandary

we arranged the million most-common words
alphabetically along the Moon's orbit

and added every possible word-group

and then we tightened each word/group's orbit
according to how common it is

and linked groups to component words
with imaginary elastic bands

and freed each word/group to slide
within its own orbit

speculating that the elastic
would pull words into groups
approximating the Yahoo toplevel categories:

Arts&Humanities, Business&Economy, Computers&Internet,
Education, Entertainment, Government, Health, News&Media,
Recreation&Sports, Reference, Regional, Science,
Social Science, Society

arranged somehow around the disk of orbits

but mathematically
the elastics would really just
clump everything into a tarball

so a subtler metric is needed


take the frequencies of any two words
and calculate the expected average distance
between occurrences of these words

calculate the actual average distance
(smaller for related words,
longer for unrelated)
and make the elastic 'repel'
when the distance is longer

but even then
the repulsion can act
'clockwise' or 'counterclockwise'

and it's not obvious how or whether
the collective force would neatly
fill the circle/orbits

so i want to virtually 'jiggle'
the system, with elastics
so that it finds its minimal energy state

(but is this realistic?)