Reprinted from https://blogs.adobe.com/accessibility/2014/09/pdf-accessibility-starts-with-the-author.html, by Matt May:
PDF accessibility is a popular topic around here. We believe access to the universe of PDF documents out in the world is critically important to users with disabilities, and we try where we can to help authors make PDFs as accessibly as possible using Adobe Acrobat and other products we make.
Freedom Scientific recently announced a beta version of their flagship screen reader, JAWS 16. In that release, their Convenient OCR (optical character recognition) feature has been expanded to help take inaccessible PDF documents (for example, paper documents taken directly from a scanner) and make them readable to screen-reader users. In light of this news, we want to caution creators of PDF content regarding what this new feature means for PDF accessibility.
Authors frequently ask us to automate this process for the thousands, millions or billions of documents they’ve already produced, and while tools do exist to do this, there’s no guarantee that they can be 100% effective at what they do. Likewise, tools like Convenient OCR, made for readers to access previously inaccessible content, all have their own limitations.
None of these tools is a silver bullet. Neither automated authoring tools nor assistive technology for readers is intended to relieve creators of PDF content from their responsibility to make that content accessible.
The PDF format, along with tools which implement the PDF/UA specification, have many features intended to enable a final PDF document that works for users on the devices they choose, with the software they choose, and with the accessibility features they prefer. No matter what technology may come along, authors know more about the structure and content of the document than any algorithm can.
The Convenient OCR feature is what’s known in accessibility as a repair technique. As with most other things, when you have to repair something, it means you’re starting with an item that is broken. This is true of inaccessible PDF documents: for example, when a page is scanned into a PDF without being properly tagged, readers miss both the textual content and the document structure that sighted users take for granted. When a tool tries to repair that document, it can only guess at the original text (and we know first-hand that OCR still isn’t 100% effective) and what kind of structure was intended.
Authors alone have the ability to communicate exactly what they mean, in a way that can be read by people with perfect vision, or who use a screen reader, or who use magnification software and/or features like high-contrast mode to adapt to their individual needs.
None of this is intended to take away from the work that Freedom Scientific is doing. Repair techniques are invaluable tools, and features like OCR have been used to work around inaccessible text since the 1970s. We feel this is a great feature for JAWS users. And other assistive technology companies such as Dolphin and AiSquared have also built tools to make text found in graphics readable to their users.
But for authors, the rules don’t change. We have lots of tools and tips on how to create accessible PDF documents that can be used by people of all kinds of abilities and preferences, and it’s still our responsibility to use them.