Writing against text recognition

Workshop proposal by Ema

In one sentence:

Collectively develop an handwriting that can be read by humans and not by OCR technologies (Optical Character Recognition).

Idea

The idea comes from the unreadability of graffiti tags. Those signs in the urban environment are very difficult or almost impossible to read by many. The only ones that usually can read those glyphs are other graffiti writers, people that are already part of that bubble. Hiding in plain sight, those letters, those signatures, those names that surrounds us all over, incoherently are actually readable (NOT recognisable) by an handful of people.

Tags and graffiti are based onto the development of letters. The development of an hand-style is at the core of the graffiti practice. Letters do not pretend to be perfect and clonable as like a font. When coming to tagging, they have to be fast to be written. Exactly like our very own calligraphy, graffiti writers develops their very own calligraphy using the same letters of our common alphabet.

So why aren't we able to read them? Why don't we develop our

Unlike this communal aspect of calligraphy being tied to ourselves, how we normally write can be now a day easily be scanned and recognised by an image recognition algorithm (commonly known as OCR - see more: https://cloud.google.com/use-cases/ocr + https://docs.cloud.google.com/document-ai/docs/ce-with-genai).

By personal experience, i noticed that Generative Algorithms really struggles to create graffiti. This is because this kind of AI is trained on images thus do not understand that besides the aesthetic of graffiti, there is the letter aspect to it. AI mimics graffiti in a very gibberish sense. Probably most of us still recognise them as graffiti, but actually

To develop a character set --a font-- is useless to the objectives of this workshop, cause a font is just a graphic look of a letter, an overlay layer to a character. Behind it there's always the correspondence to a specific Unicode character. When highlighted and copied, a digital text written on a computer, does not take into account the appearance of it. Every letter is natively encoded into the Unicode characters set, so for instance to develop a font, will be just useless cause the machine can still perfectly read what's written, no matter how complicated it looks.