Adjusting extraction confidence in field specifications

Note that this information applies to the fields used with document specifications (more information). Fields are also used when preparing invoice profiles (more information).

Use this procedure for both field specifications and field templates

1. Open the field's Extraction confidence settings.

2. Set the Confidence level by clicking and dragging the slider.

   ---

   Should you adjust this setting for a field?

   It can be useful to decrease the confidence setting if Verify operators are confirming too many fields that are correctly extracted.

   It can also be useful to increase the confidence setting when fields with extraction errors are receiving Complete status. Sometimes it helps, and sometimes it does not.

   Alternatives:

   • Set up basic validation of fields.

   • Add a validation rule to the field specification.

   • Select Confirm value, Confirm filled-in data, or Confirm found value (depending on the field type) – see below.

   ---

   The default setting for barcode fields is 5; for other types of fields it is 6.

   • Confidence level 10 is the highest level, resulting in more manual confirmations, but few mismatches.

   • Confidence levels 9–7 give you fewer manual confirmations, but possibly more mismatches.

   • Confidence level 5 or 6 is the normal/default level.

   • Confidence levels going down towards 1 result in fewer and fewer manual confirmations but more potential for mismatches.

3. If you select Enter data manually (no interpretation) data must be entered manually in the Verify activity. (The Extract activity creates the field but leaves it empty.) The manually entered data is then validated using the same settings as data extracted by the system.

   This setting is not available for all types of fields.

   • Reasons for selecting this setting.

     You might want to select Enter data manually, and not totally rely on the program, if:

     • The field contains a signature that, while it can be extracted without difficulty, requires manual verification to confirm that it is indeed a valid signature.

     • The field information is in a language, for example Japanese, that cannot currently be extracted.

     • Extraction of a specific field is problematic (for example a field containing Chinese characters).

     • A field is not always found (for example if a Chinese keyword cannot be extracted correctly).

   • Three ways to use this setting and how they each affect Verify.

     • For a structured field on a document layout: Assign a structured field to an area of the layout, specify a format string, and select Enter data manually. An image of the field is always displayed above the field edit box in Verify.

     • Specify a format string and select Enter data manually. (No keyword is specified under Location and there is no layout). During verification, only an edit box (where an operator enters the field value) is shown for the field. No image of the field is displayed.

     • Specify keyword(s) and select one or more boxes under Data in relation to keyword, specify a format string, and select Enter data manually. During Extract, the program searches for the keyword. If found, an image of the field is displayed above the field value edit box during verification. If the field is not found, the message "Field not found on document" is displayed. The Verify operator examines the document to see if the field exists or not.

4. You can select:

   • Confirm filled-in data, Confirm value, or Confirm found value (depending on the field type): Fields that are found on documents must be confirmed during the Verify activity.

   • Confirm empty, Confirm not found, or Confirm empty or not found (depending on the field type): Fields that are empty for which no value is found must be confirmed during the Verify activity.

   If you select both options, all fields must be confirmed manually.

5. Specify a value in Default field value if no data found...

   • ... to fill in a field that almost always has the same value (for example a currency field that is nearly always in US dollars).

   • ... to display a changeable, default value for a Verify operator. Use together with Confirm value or Enter data manually.

6. If you select Confirm if handwritten, a field that contains handwritten characters must be confirmed. (Not available for check box fields.)

7. Specify the Verification order to control the order in which fields are displayed and verified in Verify's Field list. More information.

   Unless specified otherwise, all fields have a verification order number of one (1). More than one fields can have the same number, and the specified numbers do not need to be consecutive.

   You can get an overview of the specified verification order numbers in the Fields on... list on the Document specification editor: Extraction tab.

8. Select Hide field if complete to keep fields that have Complete status when they reach Verify from being displayed in the Field list during production.

9. Click OK to save the new settings and close the dialog.

Related topics

Using fields to extract information from documents: Overview

Defining fields to be extracted: Overview