Last year, as part of our AIIM Conference coverage, we ran a story on Moonoia, a Belgium service bureau that had developed its own capture software, including handprint and handwriting recognition [see DIR 3/24/17]. Those recognition capabilities, which were developed using artificial intelligence and deep neural networks, were packaged under the docBrain label and brought to market as an on-premises application. This year, in late March (a couple weeks prior to AIIM 2018), Moonoia announced it was making docBrain available as a cloud platform.
“docBrain is built on a deep learning platform on which we have pretrained on several modules related to document capture,” said Wim De Maertelaere, a longtime capture industry executive and Moonoia’s business development executive. “These modules include document classification, extraction (including handprint and handwriting recognition), and fraud detection. We’ve generated RESTful APIs that enable these modules to be integrated with line of business applications.”
One of Moonoia’s success stories involves docBrain being used to capture multiple fields on healthcare claims forms in Belgium. These forms typically include a dozen handwritten or hand-printed fields (filled in by a doctor). They can also include bar codes, printed text, and official stamps and seals that all need to be captured. As a service bureau, using traditional OCR, Moonoia was scanning and processing millions of these forms per year for its client the Partena Group, an independent health insurance fund with more than a million members. Partena was outsourcing any required manual data entry to another organization.
In 2015, Moonoia proposed an AI-based approach designed to improve recognition rates, increase the automatic detection of inconsistencies, completion errors, and fraud; and decrease the manual work—all without disrupting the existing business flow. Moonoia even offered to build out the proof-of-concept at its own expense. After impressive initial results in 2016, Partena committed to going all-in on the AI-based docBrain system.
The result on more than 7.7 million forms processed in the first year of full implementation was a 70% reduction in manual labor, as well as significantly shorter average turnaround times. This resulted in a tremendous ROI. For this implementation, Moonoia was honored with a Project of the Year award from U.K.-based Document Manager magazine.
“Our primary differentiator is our ability to accurately extract handprint and handwritten data fields,” De Maertelaere told DIR. “Most organizations with requirements in that area are doing manual data entry, because they don’t know there is anything available that can work as well as docBrain.”
The system it trained by examples. “To train it on a new type of document, we recommend using about 5,000 example documents along with their truth sets,” said De Maertelaere. “Once implemented, the system is retrained with each exception, so it’s constantly improving.”
Because it doesn’t rely on any dictionaries, docBrain is language agnostic. Overall, it was used to process more than 80 million documents in 2017. “All of that was done through on premises implementations,” said De Maertelaere. “We are now rolling out docBrain in the cloud (utilizing the Google Cloud Platform) and looking for integrators worldwide to deploy it. The idea is for integrators to build their own applications with docBrain that they can test, train, and make available to other integrators and end users.”
After AIIM, De Maertelaere was headed to the Professional Association for Customer Engagement conference being held in Atlanta, where Moonoia was a sponsor. “We are looking at call centers as a potential user of our software for processing items like customer complaint letters,” he said.
Moonoia also announced a partnership with Contextor, a French RPA vendor. “With Moonoia, our clients will benefit from excellent capabilities to efficiently handle handwritten documents in their business processes,” said Pierre Col, Chief Marketing Officer for Contextor, in a press release.
For more information on Moonoia, visit http://www.moonoia.com/