Adding 'merit modeling' to one of our tasks (bitcointalk visualization and analytics).

It is great because it gives us a solid labeled dataset regarding what people like about posts. 

There is a problem in that how do you filter out 'merit farms' from actual meritous posts so we could use things like Ranking to choose our sample for modeling. 

I wonder what the model would tell us about merit givers...  I am sure there will be clusters of merit providers who are interested in different types of merit...