Table 5.
Impact of the two-pass recognition method on the test set.
| Entity type | No two-pass | With two-pass | Δ | ||||||
|---|---|---|---|---|---|---|---|---|---|
| P (%) | R (%) | F (%) | P (%) | R (%) | F (%) | P (%) | R (%) | F (%) | |
| Patient | 81.65 | 75.60 | 78.51 | 86.73 | 84.76 | 85.73 | +5.08 | +9.16 | +7.22 |
| City | 77.69 | 75.00 | 76.32 | 83.95 | 78.46 | 81.11 | +6.26 | +3.46 | +4.79 |
| Hospital | 81.29 | 70.51 | 75.52 | 80.96 | 76.80 | 78.83 | −0.33 | +6.29 | +3.31 |
| Doctor | 89.24 | 83.81 | 86.44 | 96.56 | 83.79 | 89.72 | +7.32 | −0.02 | +3.28 |
| Medicalrecord | 96.86 | 87.68 | 92.04 | 96.03 | 91.71 | 93.82 | −0.83 | +4.03 | +1.78 |
| Zip | 100.00 | 93.57 | 96.68 | 100.00 | 94.29 | 97.06 | 0.00 | +0.72 | +0.38 |
| Idnum | 89.47 | 78.46 | 83.61 | 90.00 | 78.46 | 83.84 | +0.53 | 0.00 | +0.23 |