How Zonal OCR Improves Forms Processing Accuracy

When the time comes to extract data from standard forms, simply scanning the entire document isn’t an ideal solution. This is especially true of forms that include instructional text, since you probably don’t want to keep capturing “Directions” from every form. Even when looking only at fillable information, there can be a lot of text to capture. Optical character recognition makes it simple to automate data extraction as part of a forms processing workflow, but the most effective frameworks utilize a specialized form of recognition known as zonal OCR. 

What Is Zonal OCR?

While zonal OCR still identifies machine-printed text and matches it to existing character sets before handing it off to another stage of a predetermined workflow, what sets the process apart is the way it goes about reading a document page. A typical standard form often features multiple fillable boxes where someone can enter their information. It could also include drop-down menus with predetermined responses (suffix, state, and country are all common examples of this). Trying to recognize all of that text at once greatly increases the number of possible results, which could impact both accuracy and performance.

Zonal OCR addresses this challenge by splitting the page up into several distinct zones, each of which typically corresponds to a form field (although it doesn’t have to). Instead of reading the entire page, then, the OCR engine selectively recognizes the text in these zones. It can also be combined with form image dropout, which removes text and graphical elements that don’t need to be read and might interfere with the recognition process. By reducing the amount of text that needs to be matched, zonal OCR and significantly improve recognition speed and accuracy.

Limiting Recognition Results

The most effective OCR solutions then go a step further by designating the type of information that should be found within those zones. This reduces the range of potential outcomes, which makes it easier for the OCR engine to return an accurate reading.

For example, the letter “Z” bears superficial similarities to the number “2.” If the OCR engine needs to take into account all possible responses, it may struggle to distinguish between the two accurately, especially if an unusual font was used to complete the form. However, if developers stipulate that a particular “zone” should only include numerical values, the OCR engine suddenly goes from having to consider dozens of letters and special characters to just ten numbers. This makes it much easier to obtain an accurate recognition result.

For hand-printed form responses, applying the same zonal strategy to Intelligent Character Recognition (ICR) is especially helpful. Going back to the “Z” and “2” example, the distinctions between the two characters are often much more subtle in the case of hand-printing. If a form field includes the date to be printed out in a month/day/year format, there is no reason to include a “Z” in the list of potential characters that might be found in that field because no month includes a “Z.” When the ICR engine comes across a “2,” then, it’s more likely to identify it correctly because there are fewer potential alternative characters.

By constraining possible recognition results over a smaller range of defined character sets, zonal OCR and ICR both greatly improve accuracy when it comes to forms processing. The list of potential results is typically referred to as a data validation list.

In addition to constraining character sets, regular expressions can also be applied to different zones to specify what kind of data is expected to be found there. A regular expression is simply a string pattern that sets rules for how characters are formatted, such as a phone number, Social Security number, or credit card number.

Setting Up Zonal OCR

Integrating zonal OCR capabilities into a forms processing workflow first requires the creation of specialized templates that map out the location of each field that contains data. In any organization, the various types of standard forms received should always be built as templates within the solution. This allows the application to both match incoming forms to existing templates, but also align them to ensure that everything is in the proper location. 

The alignment step is extremely important for effective data extraction. Zonal OCR is set up to read only specific areas on a document page. These zones have clear boundaries, and anything caught outside that boundary will not be read while any character that’s only partly within the field will likely return an error result of some kind.

Accusoft’s SmartZone OCR/ICR integration, for instance, works most effectively when paired with the FormFix SDK, which handles form template creation, identification, and alignment. As part of the broader FormSuite solution, these integrations are extremely effective when it comes to streamlining data capture.

Improve Data Capture Accuracy with SmartZone

With OCR and ICR support for multiple languages, SmartZone is a powerful data extraction tool that can be incorporated into an application individually or with the rest of the FormSuite collection. It provides fast, accurate text recognition on both a zonal and a full-page basis. Developers can set up expected character patterns for fields and designate different regular expressions for all of them to deliver results that are significantly more accurate.

SmartZone not only provides out-of-the box support for pre-defined character sets, such as upper and lower case characters, arithmetic symbols, and currency symbols, it also allows developers to edit those sets to improve accuracy, confidence, and speed.

Find out how the SmartZone OCR/ICR can enhance your application’s forms processing data extraction today by downloading a free trial.