Before scanners, if one wanted to make a copy of a piece of writing one had to do it manually. Then, even with scanners, one wouldn't be able to edit a scanned document and would still have to transcribe. With OCR, there is no more transcribing.
Optical character recognition (OCR) is the process of converting scanned images with text into machine-readable text. OCR technology has been around for many years, but with the advent of artificial intelligence (AI), it is now becoming more accurate and efficient than ever before.
This technology has great potential because it can be used anywhere that humans can read text (books, book covers, newspapers, magazines, advertisements) and convert these images into machine-readable text.
Thanks to AI, optical character recognition software can not only accurately convert text from images, but also identify and correct errors in that text. This makes optical character recognition a vital tool for businesses and organizations that need to digitize large volumes of documents. With the help of AI, optical character recognition is quickly becoming one of the most indispensable tools in the digital age.
Within advertising, OCR is used across the board for native ad placement, understanding the intent or meaning of images and for analyzing ads to ensure they meet the right parameters for the ad server. Let’s dive into how exactly OCR works.
First, OCR uses a scanner to process the physical form of a document. The OCR software, then typically converts the document into binary or two-color. Then, it analyzes the image for shapes that resemble characters using an algorithm.
There are two types of algorithms that OCR software can use to recognize text within an image:
OCR software sometimes uses pattern recognition to look for patterns based on examples of text it has already been given. These examples can be in a variety of fonts and formats so the software has numerous examples to refer to. The software will compare images to patterns fed to it and pick out text in images if it finds shapes that match its references.
OCR software using feature detection relies on a given set of rules for each character that enables it to recognize those characters in a document. A character's features might include the height, width, slant, and curvature of the letter M. The rules are then applied to the entire image of text to look for matches that result in words being identified.
In the third stage, AI corrects any mistakes in the resulting file. The easiest approach to do this is to train the AI on a certain lexicon of terms that may be found in the document, only allowing words/logic with appropriate meanings to ensure that no meanings are missed.
OCR is used by various ad servers to detect and audit display ads by the text or content that is within them. This can be used to make sure the ad matches the landing page content or just to ensure that there is no lewd content included in the ad. In addition, companies like Facebook and Google sometimes have a text-ratio rule for their ads that OCR helps them identify.
Facebook also uses OCR for image searching. There are theories that Facebook can use a consumer’s images to better categorize them into audiences. For example, if someone regularly posts selfies in branded t shirts, say Nike, Adidas and Reebok, they could be categorized in an affinity audience for sportswear regardless of whether their shopping data is used.
OCR can also be used in contextual advertising. When an ad server works to place an ad, it may analyze the text within the ad to match it with a website that aligns with that ad content. OCR can also be used in cases where an ad server cannot crawl a site for whatever reason to determine the content.
CatapultX is using OCR to analyze and categorize ad creative before placing it within a video. Our AI also uses OCR to read and understand characters within a video to better match creatives to a video scene or moment.
OCR will continue to strengthen. Early OCR machines read one character per minute, now they read 10K+. In 2017, OCR Solutions said that their OCR had 90% character recognition accuracy. In 2021, Google’s Cloud Platform Vision OCR tool has 98% accuracy. As contextual ads and content-based advertising become more popular in 2022, we will see OCR become a staple in how advertising servers determine ad placement and categorizing ads.
AI is evolving from a nice-to-have capability, to an everyday staple, especially in advertising. Learn everything you need to know about the future of AI as it applies to advertising.
In this FREE 15-page whitepaper, you'll learn:
- What AI is
- How AI applies to advertising
- Top uses of AI
- How contextual ads work
- How AI will impact your strategies tomorrow
- New forms of AI advertising