We spoke about Robotic Process Automation (RPA) services and how bots can be used to speed up processes and remove unintentional errors which crop up during document management. The two techniques followed in the automation industry are OCR (Optical Character Recognition) and ICR (Intelligent Character Recognition), to read any type of document and extract data.

Most industries try to maintain a singular template of documents to be scanned, but this control also does not reside with the organizations. Documents received are about 60 – 70% of the times in the expected format, these formats being printed documents in PDF/ Docx. The top OCR engines used are Google, Abby Finder and Microsoft and they have really good extraction capabilities when it comes to these formats.

 Some points to bear in mind when dealing with OCR engines:

• Most basic OCR engines compare image samples
• They are sensitive to fonts
• More advanced OCR engines uses neural networks to handle harder fonts
• Does not rely on context although it can help
• Is impacted by image and zone quality

When it comes to handwritten text the OCR method steps back and ICR takes over. Here again, all ICR engines also do not follow the same procedure in capturing the text.

Hand written documents exist in many documents across the industries.

For example:

• Physician’s prescription in healthcare
• Logistic supplied bill/invoice in manufacturing
• Check amount, Loan application forms in Banking
• Address recognition from forms in many domains.

Some automation tools have their own ICRs. The error rates of reading a document are highly proportionate with not only the information in the document but also the text format of the document.

Automating process within applications is a boon to the industries and would give way for AI in the near future.

A successful AI must start with automation, where processes are learnt by the bots to trigger logical outcomes based on inputs received.