Well in my quest to get the list of perks in a jump I have ended up trying a statistical techniques called kmeans clustering to group similar lines of text together.

Wasn't very good really, the clusters are about as good as getting all lines with a number in them, maybe a little better, but was interesting to try at least.