How predefined and unknown classification works

Applies to: All versions of Ephesoft Transact

A main part of the Transact workflow is classification. Transact algorithms learn document types and their content to correctly classify input files in their respective document type based on a user-defined minimum confidence level.

Confidence levels should be updated regularly if you have new documents coming in the input file.

Normal Classification

When the document undergoes classification, the system will try to match the new document with the learned files of all document types, determine the best confidence result, and assign your document to that particular document type.

Predefined Document Type Classification

Predefined document types are configured in your Document Assembler plugin. When a new document is uploaded, first the system will assign the document a document type by comparing it with the learned files of each document type. This is based on which document type returns the highest confidence score.

This confidence score is then matched with the confidence score of your predefined document type. If the confidence threshold of the predefined document type is greater than that of the generated document type threshold, then the document will be assigned to the predefined document type. Otherwise, the document will be assigned to the document type with the highest confidence score.

Unknown Document Type Classification

Unknown document types are those that don’t get matched with any document types’ learned files and return a confidence score of zero (0). 

In Test Classification, these document types will be classified as Unknown.

If needed, you can configure a property in the Document Assembler plugin to automatically assign unknown document types to a specific document type.