PDF Accessibility (tags, alt text, ...)

[JadedBlueEyes]

This sort of links to #114 and #129 - Typst should support accessible output formatting - for example, tagged PDF. Latex does not support this, and so people and organisations that want to produce accessible documents are forced to manually mark up documents in Adobe Acrobat.

Typst outputting accessible documents would be a huge win both for the disabled community and organisations and individuals that produce documents. I think it would make people a lot more likely to use and suggest the project as well.

For some more context, the LaTeX Project Team have been working on trying to get something like this in LaTeX for a while - https://www.pdfa.org/presentation/tagged-and-accessible-pdf-with-latex/

[ErroneousBosch]

Tagging in PDFs are a major part of the equation, but options such as being able to ensure reading order, and alt text for images/math equations also feed into it. This would be a huge deal given LaTeX's struggle with accessibility.

[hcsch]

https://www.pdfa.org/resource/deriving-html-from-pdf/ is a best practices guide for tagged PDF syntax (also covers some general information/guidance around the format), and with regards to #114 and this issue https://www.pdfa.org/resource/deriving-html-from-pdf/ is perhaps also an interesting document.
https://www.latex-project.org/news/2020/11/30/tagged-pdf-FS-study/ lists (past) news about work of the LaTeX community on this topic.

[jasonjgw]

Based on my own experience, combined with various discussions on the blindmath mailing list and elsewhere, there is general agreement that HTML output is far preferable to screen reader users than PDF of any kind, whether tagged or not. That is, for users who are blind, producing high-quality HTML output would be the best path to take (preferably with MathML included).

Most PDF reading applications don't even interpret the tags, where present, and the reading experience doesn't match what an HTML browser can do, on multiple platforms and with multiple screen reading technologies, desktop and mobile alike.

Based on the reference documentation for Typst, at the moment there are no function parameters for specifying a short description of an image - what would appear in an HTML ALT attribute. Also, the digital publishing accessibility community has developed best practices for including extended descriptions in documents. These descriptions can include entire document structures (e.g., tables, paragraphs, etc.). At the moment, the favoured technique is to use the HTML 5 summary/details elements, with some WAI-ARIA markup. There are other approaches as well. Many figures in technical and scientific documents would probably require, or at least benefit from extended descriptions, which authors should be able to include in the source markup.

Another accessibility requirement (currently a problem for LaTeX) is to distinguish between header and body cells in tables.

I'm sure there will be further issues to consider. I would be interested in testing the HTML export functionality when it is integrated into the compiler, and in contributing comments. At the moment, I use LaTeX for much of my writing, from which I generate HTML for review with a screen rader (i.e., interactive braille output and text to speech).

[u-fischer]

for example, tagged PDF. Latex does not support this

This is not true. LaTeX supports tagged PDF since quite some time. The documentation of tagpdf and the texlive vpat e.g. has been tagged with LaTeX, and there exist also some non-public implementations (e.g. from Ross Moore and from a publisher in Germany). The goal of the LaTeX Project is not to make tagging possible, but to make it automatic, free and easy to use so that an average user can handle it (and since the latest latex-dev release this already works quite well for simple documents, and it will get better until the summer).

[domoritz]

Based on my own experience, combined with various discussions on the blindmath mailing list and elsewhere, there is general agreement that HTML output is far preferable to screen reader users than PDF of any kind, whether tagged or not. That is, for users who are blind, producing high-quality HTML output would be the best path to take (preferably with MathML included).

While HTML is preferred, I just want to add that it's still important to generate accessible PDFs since some people will distribute only a PDF version and we should not exclude screen reader users from reading those documents.

[SeSodesa]

The lack of PDF/UA support is also a blocker for producing public documents, such as study materials for consumption in the European Union. This is due to the newish EU accessibility directive (link).