OCR (Optical Character Recognition)
You certainly know the annoying moment when you want to copy an important (and often long) text passage from a PDF of a document such as invoices, receipts, delivery notes, orders, contracts, etc. and copy it into another application for invoice entry or booking mask – but you can’t highlight the text! Why does this not work!?
OCR: The technicians’ point of view
At this point, the invoice or receipt for the machine consists only of individual dots (pixels). This first problem can be solved by an OCR (Optical Character Recognition). An OCR recognizes connected pixels and maps them to known characters. So, in general, classic OCR processing aims to recognize the full text from all characters recognized in an image file (e.g. jpg, png, tiff, pdf of invoices or receipts) and convert it as digital letters with meta information like position on a page, font size, fonts, etc. to be recorded digitally.
An invoice, a receipt or a document that was created with an intelligent invoice entry system, for example, returns a PDF text document that allows the text to be marked (details on PDF).
You can find more details on how OCR works and why it forms the basis for RPA and other automations in our article "What is OCR?".
OCR: The Accountant’s Perspective
Over time, an unpleasant confusion of terms has developed between accountants and software developers. Especially in accounting, bookkeepers and tax consultants, the OCR is actually an iOCR.
iOCR vs. OCR
“Intelligent Optical Character Recognition (iOCR)” can be understood as a combination of OCR and intelligent extraction of structured data, which as a result provides specific information in the form of individual data elements with semantic annotation.
The iOCR converts the “letter salad” into semantic information. That means iOCR or the intelligent OCR “understands documents” at word and sentence level. A high-quality iOCR recognizes simple things such as invoice number, invoice date, net and gross total amount or VAT on invoices and receipts. Better iOCR can then already correctly interpret sentences that can interpret e.g. discount, service period or tax codes. It is also possible, for example, to read addresses from a scanned text document – i.e. street, house number, postcode and town. Tables, such as items on invoices, receipts, orders or delivery notes, can also be recognized and recorded.
AI creates the intelligence of an iOCR or invoice capture
Through training, artificial intelligence learns how to form words and numbers from the letter salad of invoices and documents and, in a further step, to map the meaning of these words into semantic structures. In other words, to automatically recognize an indicator for an invoice number (e.g. voucher number) and to interpret the number and letter combination in its vicinity as the value of the invoice number.