http://www.adobe.com/accessibility/products.html
is a good place to start.
When a document gets written in Microsoft Word, its language is made part
of that document. If that document is later converted to a pdf file that
language information is taken in by the conversion process then becomes
the first component that starts to make screen reader accessibility of a
pdf file possible. Other components exist, but without that language
attribute presence, screen reader accessibility is disabled. Another
major requirement is ocr scanned content pictures by themselves in pdf
files are worthless to screen readers. I'm using speakup over here on
archlinux and am curious if it is now possible or some day will be
possible for me to export to a screen reader accessible pdf file I can
share with other screen reader users with no fear that the file won't be
accessible.
-- Twitter: JudeDaShiell
