Marking up changes in natural language appropriately is a WCAG 2.0, level AA, requirement (SC 3.1.2). Specifically, SC 3.1.2 states:
The human language of each passage or phrase in the content can be programmatically determined except for proper names, technical terms, words of indeterminate language, and words or phrases that have become part of the vernacular of the immediately surrounding text.
Video: PDF accessibility and the language of parts
Mange tout Rodney
Clearly, getting this right is important if the example in the video, or worse, is to be avoided (
<alert class="uk_specific_cultural_reference">Mange tout Rodney
For example, take the name John von Neumann (mathematician and game theoretician). In English language documents, most of the major assistive technologies, including JAWS, NVDA, VoiceOver, ZoomText Reader, Read&Write Gold and ClaroRead, pronounce “Neuman” as “Newman” (SuperNova and C-Pen Reader get it more or less right). But marking it up as German (in the context of an English language document) also sounds wrong. The same will go for place names. Try, for example, Antony (Île-de-France) or Bremerhaven.
Marking up language attributes is insufficient to solve all pronunciation problems. The solution to these exceptions, as well as to many other pronunciation problems lies in the adoption of pronunciation style sheets: PLS (Pronunciation Lexicon Specification) and SSML (Synthetic Speech Markup Language).
For when it really matters
Correcting pronunciation problems is a Level AAA WCAG 2.o requirement (SC 3.1.6), and as such may be a little off the radar for many. However, the early adoption of these technologies for both PDF and EPUB formats would be most welcome, especially for those working with educational documentation (particularly exam papers), technical materials, or perhaps medical literature or instructions, where clarity and ease of use are paramount, and where misunderstandings may have serious consequences.
At the time that the above video was recorded there was a bug in the current version of JAWS (version 2018) preventing it from correctly reading language attributes in a PDF (other than alt text and link text). Hence the video was recorded using JAWS 18 (the previous version). The JAWS 2018 bug has been reported to Freedom Scientific and hopefully will soon be fixed.