Our algorithm is getting there, but we need to make it as close to perfect as possible. To do this we have to compare the output from our algorithm to the rating real humans give us.