Table 10. Performance with training datasets of various sizes using automatically predicted medical concepts.
Explicit recognizer | Implicit recognizer | Micro | INC | |||||
Data Size | P | R | F | P | R | F | F | |
50 | 0.8314 | 0.7308 | 0.7779 | 0.8403 | 0.5817 | 0.6875 | 0.7568 | / |
100 | 0.8456 | 0.7527 | 0.7965 | 0.8515 | 0.6401 | 0.7308 | 0.7810 | 0.0242 |
150 | 0.8473 | 0.7628 | 0.8028 | 0.8619 | 0.6531 | 0.7431 | 0.7888 | 0.0320 |
200 | 0.8501 | 0.7736 | 0.8100 | 0.8643 | 0.6764 | 0.7589 | 0.7981 | 0.0413 |
250 | 0.8518 | 0.7774 | 0.8129 | 0.8678 | 0.6852 | 0.7658 | 0.8019 | 0.0451 |
300 | 0.8509 | 0.7796 | 0.8137 | 0.8695 | 0.6893 | 0.7690 | 0.8031 | 0.0463 |
*Inc is increment compared with baseline.