- The number of annotations for the original (non-recalibrated), processed with La Rocca (LR) and processed with our method datasets are reported in the first three columns respectively. The last three columns represent the difference between the number of annotations in the recalibrated datasets and the original dataset and between the two recalibration methods. Cases where our method results in a greater number of annotations are shaded in green, otherwise in red. In orange, the cases where the number of annotations is equal. The three datasets where we observed split annotations in LR are coloured in light blue