Editor's note: the blog post below was published in 2018. For a more recent post on optical character recognition (OCR), check out Combining OCR With AI and RPA for Advanced Data Analysis.
In our current day and age, businesses increasingly rely on digital information to streamline and scale their operations, and robotic process automation (RPA) has become invaluable to fully amplifying the potential of digital data. But despite the overarching emphasis of digitization in business practice, not all operations are fully digital yet. In fact, the BBC reports that the average UK office worker still uses up to 45 sheets of paper per day and that only around 1% of all businesses can be considered truly paperless.
Almost every industry and company department has their own instances where processes still rely on analog inputs and communication channels, with invoices, CVs, parts requests, or patient information being common examples of documents that are often processed in printed or scanned form. This is not only a problem because of logistical and environmental reasons but also because of the strain put on organizations’ ambitions to interconnect their operations and become truly digital companies.
Empowering RPA to cope with unstructured data inputs
The automation of processes involving semi-structured documents can only be achieved with a reliable digitization engine, especially because RPA software robots have to understand and interpret the different information inputs that they deal with in a rules-based fashion. When even the most basic of robots are confronted with structured data in the form of a database content or a computer-generated text document, they are already able to read out the electronic information and apply predefined rules and advanced algorithms to accomplish their automation responsibilities. However, scans of documents typically take the form of an image that does not store easy-to-read-out information on the words, fields, numbers, and graphs that it contains. In this regard, the robots are confronted with the same problem a human employee is when trying to copy and paste sentences from an image. In order to navigate these obstacles, RPA’s software robots are able to rely on integrated OCR technologies that specialize in document recognition, data capture, and language processing. This way, robots are able to deal with reading out, editing, and manipulating the information in scanned documents—just like a human would when manually typing out the information from a scan, but much more quickly and accurately.
UiPath’s OCR competencies
With UiPath’s integration of advanced OCR technologies, RPA software robots are able to interpret scans of paperwork and transfer data into digital formats in a more efficient and effective manner than ever possible before. Nonetheless, the sophistication required also depends on the degree to which the input data is structured.
Simple, structured documents
Documents that do not have much content, have good image quality, use a widely known language, and use same writing style and format are generally easier to read and understand for RPA software robots. For such cases, UiPath supports fitting solutions by Google or Microsoft that are budget-friendly and easily accessible for the required level of sophistication. If documents contain more information, have lower picture quality, use artistic fonts, or use foreign languages, extracting all the information reliably becomes more challenging and requires trustworthy integration—which is why UiPath has entered into a close partnership with ABBYY. ABBYY’s OCR product FineReader confidently deals with these aggravated conditions within the confines of the UiPath Enterprise RPA Platform, as long as the relevant information in the documents is structured according to an identifiable format.
Complex, semi-structured documents
It’s clear within any business environment that scanned documents do not always share the same structure. Invoices and CVs, for example, typically contain the same key information but can be formatted in a multitude of ways and contain information in a range of different positions. To allow companies to incorporate this kind of input into their automated processes, UiPath relies on ABBYY’s cutting-edge FlexiCapture technology, now tightly integrated into the UiPath Enterprise RPA Platform. To prepare a scanned document for UiPath’s software robots, it is passed through a series of complex steps in order to extract the information: the document is classified according to its scope, a template is created for identifying the data based on the individual classification, and the data is extracted. During this process, a monitoring tool is applied in order to get relevant analytics and make sure that the system is reliably up and running. Moreover, a human validator can be involved to check for extraction errors for future avoidance.
In providing for the transformation of unstructured data inputs into structured data that can then be further processed and communicated, such integration of OCR technologies allows for a perfectly aligned interplay with RPA. As a consequence, the content of an invoice that arrives via post, for example, does not have to be manually entered into the system. Instead, the document can be scanned at arrival, whereupon OCR can extract the relevant information and the RPA software robots are able to send the data to the company’s database.
Paper is dead. Long live paper!
Ken Weilerstein, the research vice president of market research company Gartner maintains that “[p]aper is universally accepted as valid for contracts and other legal documents, and the signatures are familiar and accepted to a greater degree than any sort of digital signature.” And, for this and other reasons, paper is likely to stay an essential part of business practices for the foreseeable future, regardless of the opportunities that digitization provides.Nonetheless, the challenges of working with printed and scanned documents often automatically disqualify many companies and their operations from being fully automated. As a consequence, resources get wasted as employees are forced into engaging in repetitive and tiresome tasks, so an intelligent combination of key technologies is necessary to utilize the full potential of automation. The integration of OCR and RPA platforms, as promoted by ABBYY and UiPath, is such a cooperation and has the potential to streamline the day-to-day processes of the majority of companies still depending on print-outs and scans.
This is a companion discussion topic for the original entry at https://www.uipath.com/blog/ai/intelligentocr-rpa-unstructured-data