Elena Brescacin

@null @fediverse I’m blind and have used OCR systems since 1990s. Computer at first, with scanners. Then camera-driven devices.
OCR products pre-llm, were very poor - it’s not the first time they detected “O” letter instead of “D”, “m” in place of “rn” and vice-versa, Llm has brought some improvements in that approach. The one I’m currently using, called “BeMyAI” inside the “Be My Eyes” desktop and mobile app, provides better information; it’s OpenAI API trained with blind people’s need and it corrects possible errors. Is it always right? No. It always needs human verification especially for sensitive data. But to read names of a food or a medicine they’re OK. Never used those, to read a paper book. It’s not worth the effort, buying paper when there’s a high risk of error, if there’s the digital.

@zqps I’m afraid discrimination has become mainstream. Against queer, against black, against people with disability…

@Quokka Sorry, and thanks for explanation. It’s a serious situation anyways. This kind of devs shouldn’t be accepted into open source community; maybe I’m too extremist

@Quokka Wait a moment. Is it the same tesseract that develops the OCR? Because I sometimes use that OCR for NVDA screen reader (non visual desktop access). If so, I’ll take it off immediately! This kind of hardcoded deny list, is an abuse. Worse than the one I noticed 25 years ago on MSN messenger where you couldn’t label yourself as “John Doe - webmaster of [project]” but then if you named yourself “I’m the drill for young asses” (said it less rude than it was), Msn messenger let you place that display name upfront.
Let me say that if you develop a software where you deliberately block pro-trans, pro-sustainability or topic you dev potentially dislike, your software shouldn’t be distributed to the public.

@writefreely @macfranc @fediverse Now waiting WriteFreely '17 into @yunohost catalog as well!