Cool experiment!

BNtw. is there any similarity measure that can be calculated with less complexity? (e.g. without the need to compare every pair of comments?)