"Sure, there was spam in the wordfreq data sources, but it was manageable and often identifiable."

How sure can we be about that?