OCR Image Processing: How to Leverage Optical Character Recognition to Work a Bit of Magic
New digital technologies are emerging all the time, but one venerable favorite continues to hold its own among the newbies—OCR, or optical character recognition. OCR’s origins date back as early as 1914, to technologies involving telegraphy and creating reading devices for the blind. Early versions needed to be trained with images of each character and worked on one font at a time. Today’s OCR is much more sophisticated, but it continues to learn and get better at what it does—every time it’s put to work. The technology is foundational to the power of the document management solution DocLink. We thought it could be both informative and helpful to share how OCR works a bit of magic, and in doing so, makes your work life easier.
How DocLink uses OCR
DocLink is a Document Management System that works hand in hand with OCR to streamline processing of images.
DocLink’s versatility in document management allows the import of captured documents from a wide variety of sources. One of those is from an OCR system. Once processed by OCR, the document is secured inside the DocLink centralized repository. The first step OCR performs is a smart scan. OCR works to intelligently recognize alphanumeric characters to save you from having to manually index each document. Using OCR, it can locate critical data from the document—data that you need to take action on the document, such as the vendor name, invoice number, invoice amount and due date. By automating this process, OCR helps you boost efficiency and reduce transaction processing costs. DocLink uses the captured values passed from OCR to help drive automation and later provides multiple ways to search and recall images. Hint: Manual indexing and storage of the documents is automated.
Imperfect—but politely so
OCR is an imperfect technology—its ability to “read” is based entirely on the quality of the source document. Not surprisingly, original documents—as opposed to scanned images—are usually of higher quality and therefore result in more accurate reads.
OCR recognition mediates the inherent imperfections by explicitly telling you what it doesn’t know. The OCR verification station lists documents that have gone through the recognition phase, where documents with data elements that couldn’t be definitively identified are earmarked for your interaction.
A lifelong learner
As you subsequently review that document, you are actively teaching the OCR engine how to get better. For example, if OCR tells you it cannot locate the Invoice Amount, you show the system where that value lies within the document, you are teaching it what to look for and where—making the next similar scan more complete.
You can even watch OCR “thinking” in action: hover over a field as you’re reviewing the document and a pop-up display will show the value it has read. You can accept the value or correct it as necessary. To be fair, OCR does its very best, after all there are multiple ways a supplier might denote a particular field. For example, the Invoice Number might appear as: Invoice #, Invoice No., Invoice: or Inv #. Fortunately, OCR has logic embedded in the OCR engine to handle most of those commonsense variations, so you won’t have to teach the system those.
We hope you’ve enjoyed this behind-the-scenes peek at how OCR works its magic.