Correcting False Negatives

A false negative occurs when FormFix fails to match an image to a template when the two should match. Poor quality images are a common cause of false negatives (for more information, see Enhance Images). Images with specks or borders may identify slightly less accurately than clean images. Distorted images, such as bad faxes, will rarely pass identification. For more information, see Handle Similar Templates, Drop-Out and Faxed Images.

Adjust IdentificationQuality

Increasing the IdentificationQuality property corrects most false negatives. Check the resolution of both the filled form and the template to verify that both are set, and set to the correct value.

Template Adjustments

• Rarely recognized - If a particular template is rarely recognized, there may be some problem with that template. Templates should never be skewed and should be free of borders and specks. If the template looks good, it may not have enough content to reliably be identified. For example, the "form" portion of a credit card slip may consist of a box and a few words. That may not provide enough content to be recognized.
• Similar Templates - If a filled image matches a template with a high confidence (greater than 80), but still fails to identify, there is a high probability that another template in your form set is very similar to the first template. The IdentificationProcessor may have trouble distinguishing between these similar forms. If this is the case, the IdentificationResult will have a state of MultipleMatchesFound. See Handling Similar Templates for more information.
• Low confidence - If filled images consistently match a template with a low confidence (less than 70), it is very likely that those filled images represent a variant that should be added to the form set. For more information, see Handling Similar Templates.

Correcting False Positives

A false positive occurs when FormFix matches an image to a template, but the two are different and should not actually match.

There are two types of false positives.

• The first is a filled image that should not have matched any templates.
  If a false positive should not have matched any templates, reduce IdentificationCertainty or increase IdentificationQuality. If thes confidence for the false positive is low (less than about 75), you may want to increase MinimumIdentificationConfidence.

• The second is a filled image that should have matched a different template.
  

  If a false positive should have matched a different template, increase IdentificationQuality. If the false positive is due to the presence of nearly identical templates, see the section on Handling Similar Templates.