How accurate are Amazon's and Yelp's fake review detection algorithms? * How many false positives are there? How many businesses do they unfairly penalize (and

How accurate are Amazon's and Yelp's fake review detection algorithms? * How many false positives are there? How many businesses do they unfairly penalize (and how many reviewers do they humiliate)? For example, perhaps some behavior trait common to a location or social class is misinterpreted as a fake review. * How many false negatives are there? How many fake reviews are given extra legitimacy in the eyes of users because the algorithm gave them a pass? Maybe the algorithm weeds out only the bad fake reviewers, or a certain kind of fake reviewers. Perhaps a cottage industry of techniques to game the algorithms will arise, like SEO. I do know that decent shills know how to fool amatuer slueths; heck, I could do it: Write something long, with specifics, add a little balance to look reasonable, mirror what some others say, etc. I'd be interested to know how Yelp and Amazon can test their algorithms. They would need a large set of proven fake reviews, and a representative set would be much more valuable. Where does one find those?