Metrics policies

Platform has the following policies for calculating the overall metrics for a set of documents, that are the metrics for an experiment:

• Micro Average metric: the matches of all the documents are summed, thus producing the summations for TP, FP, FN.
• Macro Average metric: each metric is given by the average of the values ​​of the micro metrics on the individual resources (for example, macro precision is the average of the micro precision of all classes / categories).
• Sample Average metric: each metric is given by the average of the metrics on the respective documents (excluding those with undefined metrics).
• Weighted Average metric: each metric is given by the weighted average of the metrics on the respective documents, where the weight of each document is given by its number of annotations (as before, it is necessary to exclude the documents with the undefined metric).