Teaching a Robot to Read

By

Many businesses are struggling to become more efficient and drive higher levels of employee engagement and customer satisfaction.  Intelligent Automation solutions could address all of those.  UDig can help you determine if it’s right for your organization, and if it is, you may get the opportunity to teach a robot how to read.

You might think that teaching a robot to read is a crazy idea.  Why would we teach a robot to read?  Why would you try?  I should probably explain. 

Robotic Process Automation (RPA) uses software to automate business processes.  It is designed to emulate and replicate human tasks that are mundane, rulebased and repeatable.  There are hundreds of processes that make sense to automate using RPA.  What about processes that require data to be read from a document and then entered into another application, or even many applicationsShould those be excluded from the list of potential candidates for automation? The answer is no. 

What if we could automate a process and, within that process, read a document, extract the relevant information, and continue with data entry and further processing of that information?   Suppose we have a document that contains structured data. In that caseit is relatively straightforward to identify and tag the data’s location. The robot will repetitively extract that information from the same place on the same form.  That solves that problem, right?  Well, sort of, but that only covers a small percentage of the documents out there.  

Most of our clients have many different document formats, and the extraction of the information from those documents may or may not be in the same location every time. That’s where “teaching a robot to read” and Intelligent Automation come in.  By introducing document intelligence into the automation, we can teach that bot to read. 

Document intelligence combines three main technologies:  

  1. Optical Character Recognition (OCR) takes a digital image and recognizes the characters within that image for either export or use in other applications. OCR is certainly not a new technology but one that’s critical to this process.  It does require some trial and error to pick the best OCR engine for the particular use case.  There may be a combination of handwritten and typed documents that need to be ingested, and that’s where it’s good to test out a couple of engines to find the most accurate. 
  2. Natural Language Processing (NLP) allows a computer to ‘understand’ the contents of a document, including contextual nuances. NLP is important for unstructured or semi-structured data when you may need to interpret the data in a field or a location of a document for it to be extracted accurately as part of the process. 
  3. Artificial Intelligence (AI)/Machine Learning (ML) involves computer algorithms that use sample data and human training to allow for the computer to make predictions or decisions based on its ‘knowledge. Several ML models can be used for specific use cases, i.e., reading an invoice or a receipt.  These models still need to be trained for greater accuracy and require a subject matter expert (a human) to help train the model or robot to read the information.  This is where you teach a robot how to read.   

A good use case for this technology is in accounting, where you may be dealing with a volume of invoices of all format types.  You may need to identify some key fields to match up to a purchase order in your accounting system.  This can be a fairly tedious process for a human and is prone to manual errors.  In this case, you can teach the robot how to read the invoices and then have the robot enter the necessary information into your accounting system for payment.  It sounds much simpler than it is, and it’s not yet perfect, but if the robot isn’t confident about what it’s reading, it will ask a human for assistance.  At that point, the humans are only handling the exceptions and are free to do more valuable work for the company.   

What does this mean for you?  As you look to transform how work is done, these technologies improve your team’s work-life balance and speed up a modernization or digital transformation initiative.  

At UDig, wadvise and assist our clients in determining the right technology for their specific business challenges.  We implement Intelligent automation technologies to improve data entry accuracy, reduce cycle and processing times and leave your clients and employees better. 

If you’d like to discuss how Intelligent Automation fits within your organization, please feel free to connect with me on LinkedIn.  

 

About The Author

Chris Lara is the Senior Director of Intelligent Automation. He is a 30-year technology industry veteran working with clients of all sizes with a local to global reach. Chris is passionate about utilizing technology to drive value to all business functions and employee and customer experience.