Fine-tuning invoice extraction

XXX This is ok for now, but should probably be adjusted when we have more time.

Below is a list of things that you can try adjusting to improve invoice identification and extraction. The items in the list are some of the many ways of improving system performance.

Strongly recommended: Use image files for testing.

Adjusting the system setup, invoice profile, and invoice definition

• Optimize the invoice definition.

• Improve buyer identification.

• Adjust the identification settings of the invoice definition.

• Connect fields to business logic.

• If the system has problems extracting a field, improving the format specification can greatly improve field extraction. If a field has a well-defined format that is the same on all invoices (or on all invoices from a single supplier), define the format specification precisely.

  Whenever possible, avoid using X in the format string to specify a character position. X means that the character can be either a letter or a number, and this uncertainty can hinder extraction. If it is always one or the other, use A or N instead of X. (See Syntax for format strings in field definitions, field profiles, and identifiers.)

• Use active validation whenever possible.

• Try applying background filters.

• Ensure that the character set of the language you are extracting contains any non-alphanumeric characters (such as parentheses) that are needed. (Instructions.)

• Adjust Eiglobalextra.ini, for example by adding customized field types or adjusting the business logic.