Figure 4.

Support tools for the validation of filters. To the left of each sentence is shown the target interaction (either from gold standard or derived by the system). Green means that the interaction detected by the system matches an interaction in the gold standard. Gold marks an interaction in the gold standard not detected by the system. Red denotes an interaction detected by the system, but not contained in the gold standard. In other words, true positives are in green, false positives are in red, and false negatives are in gold.