Quill Digital

So, you have an RPA or other bespoke software doing amazing things. How about your data? Most organisations don't have a nice and neat way of storing all their data, especially if that data comes from an unstructured source like say, a scanned document.

The new technology that Sonix have developed will allow you to read a document, and structure that data, no matter if it is different each time or the scan is not brilliant.

We achieve this by doing something very unique.

Our algorithms do not rely on OCR (Optical Character Recognition). Instead we use a different method.

One of the first projects that our developers worked on was Y2K and they were tasked with deploying some OCR technology to read loan applications and digitise them.

Back then, a bank branch would type the application up, scan and send it by email to a central office.

There were two plans to get this stored digitally. The first was to employ a typing pool to enter the information on the banking systems or, use OCR to digitise the data. The bank went with the second option.

The software used at the time was cutting edge and very expensive. The project failed completely due to the fact that the most accurate read was only 65%. In a loan application, this was not acceptable.

So the typing pool option was revisited and the developers moved onto java applications.

Many years later while sitting on the tube in London, a gentleman got on with his copy of the Times cryptic crossword. Having noticed that only 65% of the crossword was completed, a probability model to mimic this was formed within Sonix.

In its very basic form, this is essentially how Quill works. It can take a variety of scan types (pdf,jpg etc.) and accept the 65% accuracy read, but then using a machine learning model to supplement the remaining 35% with the corrected words.

Find out more

Installation Guide
OCR Settings
Replace Words Settings
Orphan Words