There isn't a definitive way of telling which transactions are created with the intention to spam or not. Any statistics out there is a hunch at best. Even if you do implement this, you will have to get clients to reject blocks with those transaction in them since they are still considered as valid, albeit non-standard. It isn't a solution at all.
we can't exclude any transactions in Bitcoin or that would be considered as censoring, no matter what your agenda is.
what is your opinion?