[BlindMath] Questions about Nemeth and UEB

[Neil Soiffer]

A comment, a shameless plug, and a request for feedback...

Comment: Both JAWS and VoiceOver/Safari only support Nemeth from MathML at
the moment.

Shameless plug: If you want UEB output from MathML, you need to use
NVDA+MathCAT.

Comment + shameless plug: If you would like to compare Nemeth and UEB, you
can do so by looking at the MathCATDemo page
<https://nsoiffer.github.io/MathCATDemo/>. That page is meant to show what
accessible math can be like. It is NOT meant for people who need to read
math -- use NVDA+MathCAT on your web page or Word document for doing actual
reading. The demo page is fully accessible with the braille text area
marked as a region so it shouldn't be too painful in a screen reader to
jump between input and braille. The input is pretty flexible -- it takes
ASCIIMath <http://asciimath.org/>, TeX, and MathML as input. You can turn
off the speech if it is annoying (and it will save me a few tenths of a
penny also). Besides Nemeth's use of drop numbers, the most obvious
difference between Nemeth and UEB you'll see is that UEB is longer,
sometimes considerably so.

Request for Feedback: if you find a bug in the braille output (either
Nemeth or UEB), PLEASE let me know. The quality of braille people accept
from screen readers is shockingly low. I am striving to make MathCAT an
exception -- tests I have done seem to bear that out. If you are a UEB
reader, then you know there are grade 1 symbol, word, and passage
indicators. Both BANA and ICEB have vague guidelines on when to use which
for math... and they contradict in a few cases. I'd like to improve
MathCAT's heuristics for that, so if you feel MathCAT should have chosen a
different one from what it did, please send me the input and an explanation
of why you think the different indicator makes sense in this case. Bonus
points for trying to come up with a general rule.

Finally, if you have suggestions for options to control UEB output, let me
know. I am NOT a braille reader, so some things I find surprising in
braille may seem very sensible to someone who reads braille. For example,
"sin x" is " ⠎⠔⠀⠰⠭" but "2sin x" is " ⠼⠃⠎⠊⠝⠀⠰⠭" -- the contracted "in" in
the first expression is now spelled out because the '2' introduces grade
one mode. To me, that difference makes it harder to quickly recognize
"sin", and hence patterns in the output containing "sin". Maybe these
differences are second nature to someone who knows UEB well and it doesn't
get in the way of recognizing patterns. A previous poster recommended using
uncontracted braille. If that is something you'd find useful, please send
me email (I don't think the mailing list needs to know everyone's thoughts)
and if several people think that would be useful, I'll add an option to
MathCAT to force grade1. Other suggestions for improvements to MathCAT are
also welcome.

Neil Soiffer