zlacker

It was a bad approach to the problem.

>>eli_go+(OP)
Perhaps it could be made into a good approach if it were reworked in terms of topic modeling or clustering categories and them feeding them through a machine learning KNN or bayesian model enabling prediction.