The top figure contains raw counts for the number of song titles from my iTunes library that contain the anatomical words listed on the abscissa. The bottom figure is the the same figure, but this time normalized using the sum over all song titles with some piece of anatomy in their title. You can think of this as the conditional probability of a specific body part, given that we already know the title contains some sort of anatomy. So, for example, Probability( head | category of word is anatomy ) ~ 0.19. I didn't bother calculating the joint probability, because I'm lazy. In that case, we would just multiply: Probability of ( anatomy_label & category of word is anatomy ) = P(anatomy_label | category of word is anatomy )*P(category of word is anatomy ).
Fuck. I should really be doing homework.
This is getting me nowhere.
Also, let me know if you can think of any labels I missed.
2 comments:
My iTunes would turn up several instances of the word 'cunt'.
How did you get your computer to understand (i.e. compute) lyrics?
i grouped according to whether word X was contained in the song's title.
Post a Comment