Reverse OCR is text recognition for testing.
Optical character recognition (OCR) is not a new technology. Some vendors even claim they have a brand-new AI based OCR. OCR has always used algorithms that are classified as artificial intelligence, today. In testing you do not stop at OCR you need Reverse OCR.
The shortcomings of OCR
The accuracy of OCR is mostly measured against prints. With an accuracy of 98 -99 percent per 1000 characters you have roughly 980 to 990 characters correct. In most cases this level of accuracy is acceptable.
In testing you merely check for 1000 characters; you mostly check a label or some dynamic data field. What happens if those 10 to 20 wrong characters are exactly within those few fields you are checking? Suddenly, this 99 percent accuracy drops to 50 percent accuracy. It drops even more if your not checking for well-known words that the OCR can identify in its dictionary.
Reverse OCR addresses these problems
In testing, OCR should be used for verification, not for reading text. Reading text is what OCR engines usually do but that is a misconception for testing. In testing you have expected values and need to verify that your subject under test provides that value. This is what we call the Reverse Text Situation: The text you want to read is well-known before you try to read it. Instead of only having characters you need to identify you have to sources, the character you need to identify and the expected result. Based on this, TestResults.io uses Reverse OCR. A uniquely available technology that builds on those two sources. TestResults.io uses different OCR engines to identify a set of potential characters and their combination. Based on your input we calculate a confidence score between what you are looking for and what is on the screen. This allows you to be 100 percent certain about your application output.
Try for Free
Book your FREE demo now!
This reduces flakiness
As any OCR engines uses AI the outcome is not 100 percent predictable. A single different pixel might lead to a completely different result of the OCR. With Reverse OCR those shortcomings are reduced to zero. This means an increased stability of your test cases out of the box.