For the new automatically learned classes the system uses fixed and pre-determined engines that cannot be adjusted. Even though a new learned DocClass has pre-defined engines that cannot be adjusted, the classes used in Supervised Learning can have any other engines assigned. When new classes are created automatically, they initially have the configurations of the engines mentioned below. However, the configuration can be changed, adjusted, or tuned manually.
The Supervised Learning Workflow uses the following engines.
- Brainware Layout Classification to automatically train level 1 classification - usually called vendor-classification in the context of BIC projects used for invoice processing.
- The Associative Search Engine to automatically detect a new class name, and, optionally, to automatically train level 1 classification.
- Brainware Extraction to train automatic extraction of header fields for fields that have Brainware Extraction and Format Analysis engines assigned for the parent base document class.
- Brainware Table Extraction to train automatic extraction of header fields for fields that have Brainware Table Extraction engine assigned for the parent base document class.