The Brainware Field Extraction engine introduces a dimension in automatic extraction of header field data on a generic level, that means when multiple logical document classes are handled with a single internal knowledge base. The engine uses an advanced N-gram Triton technology and provides high quality field extraction based on the following statistical evaluation.
- Surrounding keywords around the learned field
- Format of the learned field
- Typical location of the learned field
- Correlation between different fields learned in context of one document class.
All the information above is learned automatically. The engine also autonomously extracts candidates, which avoids time consuming specification of regular expressions through the Format Analysis Engine.
The engine fully supports the Supervised Learning Workflow.
You can configure and train the Brainware Field Extraction (BFE) engine in the same way as the Brainware Extraction engine. For more information, see Configure the Brainware Extraction Engine.
You can assign the engine to any header field of a document class using the Evaluation tab of the field definition settings and selecting Brainware Field Extraction from the list of available engines.
As for the Brainware Extraction engine the BFE settings view also displays the learn status for the selected field on the General tab.