Topic modeling Budget 2016

I topic modeled all the Estimates of Appropriations documents from the NZ 2016 budget. Don't ask me why. I used Latent Dirichlet Allocation to build 20 sets of words that each match with a "topic" in the text* of the budget. Then I made the words of each topic into a word cloud, because it's still 2008 and word clouds are cool.

Thanks to @garibaldu for the LDA code.

*The allocations are really about the numbers not the text, but so what.