Table 5.
Test data results broken down by semantic type with requirement of accurate negation, including spans of discontinuous text, with strict and relaxed matching expectations
Scored items | Matching standard | TP | FP | FN | Precision | Recall | F1 | |
---|---|---|---|---|---|---|---|---|
Disorder | 112 | Strict | 80 | 23 | 20 | 0.78 | 0.8 | 0.79 |
Relaxed | 87 | 24 | 12 | 0.78 | 0.88 | 0.83 | ||
Family history | 5 | Strict | 5 | 0 | 0 | 1.00 | 1.00 | 1.00 |
Relaxed | 5 | 0 | 0 | 1.00 | 1.00 | 1.00 | ||
Finding | 70 | Strict | 47 | 7 | 26 | 0.89 | 0.65 | 0.75 |
Relaxed | 47 | 8 | 25 | 0.87 | 0.66 | 0.75 | ||
History of procedure | 15 | Strict | 5 | 3 | 11 | 0.63 | 0.31 | 0.42 |
Relaxed | 5 | 3 | 11 | 0.63 | 0.31 | 0.42 | ||
Situation affecting health | 12 | Strict | 8 | 3 | 5 | 0.73 | 0.62 | 0.67 |
Relaxed | 8 | 3 | 5 | 0.73 | 0.62 | 0.67 | ||
Disorder + finding | 182 | Strict | 127 | 30 | 46 | 0.81 | 0.74 | 0.77 |
Relaxed | 134 | 32 | 37 | 0.81 | 0.79 | 0.80 |