Table 3. Relative validation results, using the document as text window.
Subset | 25% entities validated | 50% entities validated | 75% entities validated | ||||
Measure | Method | TP Ratio | P Increase | TP Ratio | P Increase | TP Ratio | P Increase |
2*SimGIC | Dict | 35.0 | 39.9 | 68.8 | 38.7 | 92.4 | 23.5 |
CRF | 28.5 | 14.1 | 58.1 | 16.3 | 78.6 | 4.7 | |
2*SimUI | Dict | 31.4 | 23.5 | 61.4 | 21.4 | 88.7 | 15.6 |
CRF | 27.9 | 11.2 | 55.0 | 10.0 | 79.1 | 6.3 | |
2*Resnik | Dict | 31.8 | 25.1 | 73.6 | 46.1 | 96.5 | 28.4 |
CRF | 30.3 | 22.8 | 55.0 | 9.4 | 83.0 | 10.7 |
For selected subsets of validated entities corresponding to 25%, 50% and 75% of the total amount on annotations performed by each tool (Method), we present the True Positive ratio (TP Ratio), which corresponds to the percentage of True Positives remaining in the subset, and the Precision increase relative to the Precision for the total amount on annotations (P Increase). For this evaluation was used the document-wide as text window.