Skip to main content

Table 2 Per-relation and per-instance performance evaluation results

From: On the efficacy of per-relation basis performance evaluation for PPI extraction and a high-precision rule-based approach

Min #Inst. per Relation #Uniq. Relations #Instances Methods Per-Relation Per-Instance
  Pos Neg Pos Neg   TP FP Precision Recall F-Score TP FP Precision Recall F-Score
1 618 2,312 1,000 4,834 ours 132 7 0.95 0.214 0.349 153 10 0.939 0.153 0.263
      hybrid [6] 477 344 0.581 0.772 0.663 694 584 0.5431 0.694 1 0.609 1
      SST-PT [20] 195 102 0.657 0.316 0.427 217 170 0.5612 0.2172 0.3132
2 197 695 579 2,827 ours 80 3 0.964 0.406 0.571 101 5 0.953 0.174 0.294
      hybrid [6] 183 164 0.527 0.929 0.673 389 327 0.543 0.672 0.601
      SST-PT [20] 103 71 0.592 0.523 0.555 131 113 0.537 0.226 0.318
3 89 314 363 1,835 ours 51 2 0.962 0.573 0.718 67 3 0.957 0.185 0.310
      hybrid [6] 88 98 0.473 0.989 0.640 250 236 0.514 0.689 0.589
      SST-PT [20] 56 47 0.544 0.629 0.583 85 79 0.518 0.234 0.322
4 51 181 249 1,349 ours 35 2 0.946 0.686 0.795 50 3 0.943 0.201 0.331
      hybrid [6] 49 67 0.422 0.961 0.587 160 170 0.485 0.643 0.553
      SST-PT [20] 35 30 0.538 0.686 0.603 62 54 0.534 0.249 0.340
5 28 107 157 984 ours 23 1 0.958 0.821 0.884 37 1 0.974 0.236 0.380
      hybrid [6] 28 44 0.389 1 0.560 105 105 0.500 0.669 0.572
      SST-PT [20] 21 15 0.583 0.75 0.656 37 30 0.552 0.236 0.331