Sample documents

You can add sample documents for each document type in your solution. There are two types of sample documents that enable you to configure and test your solution on the Documents menu and extraction on the Fields menu.

  • Training documents

  • Test documents

We recommend two to five training documents. You can move documents between training and test documents as needed.

If you modify the training documents or the document type after training your solution it is necessary to re-train your solution. You can add additional test documents without needing to train your solution.

All sample documents should have the following characteristics:

Characteristic

Training documents

Test documents

Supported file types

TIFF, JPEG, PNG, and PDF

TIFF, JPEG, PNG, and PDF

File quality

Ideal examples that are readable with high resolution and no scan artifacts.

All expected fields should have values.

Include files that are readable with a high resolution.

Realistic scan artifacts or empty fields are useful for testing.

Unique samples

Include only samples of the same type in a document type. The samples do not need to have the same layout, only the same type.

For example, bank statements differ significantly between banks, but all included samples must be bank statements.

Include samples of any type to test both classification and extraction.

Document length

Avoid long documents with more than four pages.

Include documents of any length for testing.

Color

Avoid full-color images. We recommend bitonal images.

Include both bitonal and color documents for testing.

On the Upload training documents, the upload status of each sample document is indicated by one the following:

  • : A document is successfully uploaded and recognized.

  • : An exception occurred during document upload. This can occur due to failed recognition, an issue with Tungsten TotalAgility, or a timed-out user session.