‘Label’ represent a class (aka tag) that a data record can be assigned to. In a task, all the labels are independent and kept in a flat structure. However, they can be organized in a hierarchy for visualization purposes.
A label is applicable to all the data records of the task unless it has a label criteria. The label criteria works like a filter to restrict the set of data records the label can be assigned to (the criteria is verified). The label criteria is composed of a data record field and value of the field. For instance, a task may contain all the customer support tickets for each of the ten models of a laptop brand, and there may be two labels - "Product Registration" and ‘Power Fault’. ‘Product Registration’ may be applicable to all models, whereas ‘Power Fault’ may have a label criteria meaning it is only applicable to (for example) model B.
Performance
The performance of each label is a core feature of the system, allowing you to calculate trust in a label as well as improve the label performance. There are three metrics to measure of performance of an individual label;
Recall
<Andre to complete>
Precision
<Andre to complete>
F-Score
<Andre to complete>