Using AI to Capture a Dog Food Meal Plan or a Mobile Workers Form

FileHold can use Artificial Intelligence to help any organization get complex, unstructured documents into a machine readable structured, searchable, secure library.

AI is a valuable tool in the interpretation / rendering of the content of documents.  This article describes an extreme example of how FileHold can use this technology and a more realistic example for the mobile workforce.

The CTO of FileHold, Russ, is a dog enthusiast and recently one of his dogs was pregnant.  In such a case it is important to carefully manage the dog’s diet (he carefully manages the diet of all his dogs).  To make certain it was done correctly Russ stuck a note of ingredients, times, name of dog and quantities on his fridge as a reminder for himself and any of his family who might be doing the feeding. 

Russ took a picture of the note page on the fridge (below).  It is poorly written, has absolutely no format, no structure and is almost unreadable scrawl.  It would be extremely challenging even for a human to try and understand the text other than family members who may be familiar with the handwriting.

To demonstrate the power of AI, Russ employed an AI engine that would be a front end to FileHold.

The first line is the name of the dog “Nugget” and the quantities to feed him.  All have been perfectly rendered by the AI engine. There are some errors, for example for the next dog “Sherman” the AI interpreted this as “Sterman” still very close but understandably misread.  If the AI had prior exposure to the name “Sherman” or if it was associated with a database look up, it would probably have been rendered correctly.  The very bottom line “Green & Grey on Thurs” has also been perfectly interpreted.  All in all, very impressive.

FileHold document management software (DMS) could now use this information to process the document.  It is possible to determine what kind of document this is (dog menu) capture important metadata (dog name) and make it full text searchable for easy recovery when needed (search by date). 

Happy to report a very happy ending; healthy happy mother and puppies - nonchalant father.

As a more practical example we have a small form (below) that is typical of one that might be filled out by a mobile worker.  In this case it is a fuel inspection form that might be used in the aviation industry.  The AI engine has discreetly identified each of the fields in the form and rendered the results. 

Both the pre-existing text fields and the handwritten notes have been identified and interpreted into a more readable form.  All the typed information has been rendered by AI 100% correctly.  More impressive is the capture of the handwritten date and the numbers beside and below the invoice text and throughout the document, all are very close to rendering the content of the written form correctly.

AI helps solve an important historical problem of getting documents into a structured and secure library. Once in the library the DMS can control who has access to information and even initiate workflow processes on the documents.  A DMS keeps updated version control and can schedule documents for eventual deposition all of which is an important step in realization of the paperless office.

If you would like to learn more about FileHold contact-sales