You do not have to manually create the derived document classes, the Supervised Learning Workflow creates them automatically when it adds the documents to the learnset.
When adding the documents to the learnset, the SLW automatically creates a new derived document class. The new classname corresponds to the settings established for the document-level settings of the field Classname Format on the Analysis Editor. Therefore, this must be defined in a way that ensures uniqueness of every classname that may be created, that means that the ID, or any other unique field, must be part of the Classname Format.
Though not obligatory, it can be useful to train some additional fields for the generic class. that means that extraction results are available for the new class, and the Verification form is therefore already be partially populated.
To train the generic Document Class, switch to Train Mode. If there is only one generic Document Class, you can set it as the default classification result. When this option is set, all documents are classified to the generic Document Class and you only need to check the extraction results.