Automated Forms Recognition - Observe, Adapt, Overcome

Marketing Team - Wednesday, May 28, 2014

A major Northern California County engaged Western Integrated Systems to streamline the flow of client applications and supporting documents and reduce cycle times for paper documentation. Our investigation revealed that the county employees were struggling with the complexity of managing over 1600 different types of documents in multiple languages. Inefficient mail handling procedures had also proliferated, due to the cost and difficulty of making changes to the old data capture process. The result was a slow and painful data capture process that became less relevant every year.

External forces were also at work. The County staff were also facing unpredictable work volume increases resulting from Healthcare Reform changes and a reoccurring seasonal influx of migrant farm workers. The County needed a data capture system that was not merely accurate, but easily adaptable.

The County chose to update their old data capture process to replace Separator Sheet based, manual document classification with a solution utilizing Kofax Transformation Modules (KTM). They would use KTM to auto-recognize the form types and make the determination automatically of each specific form within a batch of case documents and index them accordingly, without the need to buy pre-printed separator sheets and without the need for manual intervention.

To do this using typical imaging software would be a mammoth task and the resulting process would be out-of-date before it was even deployed. It would have been built on classic deterministic, logic-based software and making any changes would be costly and difficult to both develop and test. Luckily, learning software like KTM produces probabilistic results using training data. KTM adapts based on experiences to enable effective responses to changes which cannot be predicted and planned for. It is very much like the way the human brain adapts to change.

Because Probabilistic Learning Software is changed via training, no big project effort is needed to update it. When new forms arrive, KTM learns to recognize them in the future. Admins simply define how the system is to respond when it sees a document type again. As new "lessons" are added, the system gets more and more accurate. Internal document experts provide the raw material that makes KTM's automatic classification system perform better and better over time. In doing so, they pass on their capability to identify key process documents to KTM. KTM then automates. That raw material consists of good representative samples of documents that are not being auto-classified correctly.

As the number of document types increases in many business processes, it can quickly become impossible for workers to keep up with the changes. However, the County’s new KTM can easily manage a vast number of rapidly changing document types in an assortment of languages. Like most learning environments, the County’s EBSD classification system depends on observations, leading to feedback and testing... that drives decisions which result in action. This is critical because rules don’t learn… and logic doesn’t adapt easily to change.

Thomas Jones

Senior Technical Engineer

Western Integrated Systems



Post has no comments.
Post a Comment

Captcha Image

Trackback Link
Post has no trackbacks.

Recent Posts