Comment on Welp that answers a lot of why all .ml are down

• condenser@lemdro.id ⁨1⁩ ⁨year⁩ ago

  How hard is it to read a font from a text?

  My man, you just don’t know how crappy OCR can be with non-latin alphabet writing systems, especially Chinese characters.

  If the source is already in text (perfectly accessible), why should we make an image out of it? That’s like saying let’s email a document, but instead of the original doc file, let’s print them out, scan, and then send the pdf of those images instead.

  source

  • whoamibro@lemmy.world ⁨1⁩ ⁨year⁩ ago

    If the source is already in text (perfectly accessible), why should we make an image out of it? That’s like saying let’s email a document, but instead of the original doc file, let’s print them out, scan, and then send the pdf of those images instead.

    That is not a correct analogy because printing and scanning a document is less convenient than just forwarding the email. But here, most people are comfortable taking a ss and share it. That’s what they’re learnt. So they keep doing that.

    My man, you just don’t know how crappy OCR can be with non-latin alphabet writing systems, especially Chinese characters.

    That’s why the OCR tools have to be improved. They should atleast be able to read the top 10 most used fonts in a language without issues.

    source

    • condenser@lemdro.id ⁨1⁩ ⁨year⁩ ago

      The analogy is just an attempt at explaining, no need to argue about them. The main point is about giving preference to the original copy, not the lossy and inefficient copy. No matter how image to text conversion tools get better there always be a gap.

      source