Sure, I'll change it around to have the option to use instantiated PSpace.

As for relying on probabilities, check out Naive Bayes. Hidden Markov Models also rely on probabilities, rather than raw counts.

Thanks for the feedback!