before precision=0.928 recall=0.806 (tp=129 fp=10 fn=31 of 160 positives)  with_website=0 with_category=0
         golden: {'tp': 53, 'fp': 0, 'fn': 0}
         variants: {'tp': 49, 'fp': 0, 'fn': 0}
         held_out: {'tp': 27, 'fp': 0, 'fn': 31}
         tricky_negatives: {'tp': 0, 'fp': 10, 'fn': 0}
after  precision=1.0 recall=0.994 (tp=159 fp=0 fn=1 of 160 positives)  with_website=159 with_category=155 category_agrees=50/50
         golden: {'tp': 53, 'fp': 0, 'fn': 0}
         variants: {'tp': 49, 'fp': 0, 'fn': 0}
         held_out: {'tp': 57, 'fp': 0, 'fn': 1}
         tricky_negatives: {'tp': 0, 'fp': 0, 'fn': 0}
