NLP needs better design

Currently only 'passing' tweets get sent to slack for review.  Any tweets that get marked as spam do not get passed on for review.

There needs to be a way to fix the false negative and the false positives.  There is only a way to fix false positives right now.