RE: [EXTERNAL] Re: [www-math] <none>

I decided all of the U+2200 block since it was introduced in the early 1990s, the math alphanumerics, plus other operators that appear in our tests and in physics. A lot of symbols in the U+2A00 block look rare to my eye, although I’m happy that U+2A2F (⨯) was added for cross product. I didn’t want to tax the translators too much. Ideally, all math symbols should be included, but I think my list includes all the common ones.

Murray

From: Neil Soiffer <soiffer@alum.mit.edu>
Sent: Friday, February 10, 2023 3:27 PM
To: Murray Sargent <murrays@exchange.microsoft.com>
Cc: Deyan Ginev <deyan.ginev@gmail.com>; www-math@w3.org
Subject: Re: [EXTERNAL] Re: [www-math] <none>

@Murray: Apropos to Deyan's question:

How did you decide which Unicode characters to support (in English and elsewhere)? Which notations? Were the decisions based on some data or (as with me) past experience/judgement under the limitations of time?

    Neil


On Fri, Feb 10, 2023 at 2:56 PM Murray Sargent <murrays@exchange.microsoft.com<mailto:murrays@exchange.microsoft.com>> wrote:
There’s also the Narrator AT, which uses the math speech engine that I wrote. Admittedly, it doesn’t know about intervals, per se, but it knows vector calculus 😊 And it can speak math in over 18 languages. As do other ATs, it says “squared”, “cubed”, “square root”, “cube root”, etc. The Unicode math symbols that it can speak are documented in Math Speech Strings and Localization - Math in Office (microsoft.com)<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdevblogs.microsoft.com%2Fmath-in-office%2Fmath-speech-strings-and-localization%2F&data=05%7C01%7Cmurrays%40exchange.microsoft.com%7C853338add7794752ceaf08db0bbe6383%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C638116684718708263%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=yaxkuMrWg5BF32lKp4FImT5oEo%2FqF0SMnf%2B3PsQYMtw%3D&reserved=0>. I hope to augment the list when we have figured out our intents.

Part of the reason I emphasize using Unicode math symbols where available instead of intent is that it’ll make enhancing UnicodeMath<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.unicode.org%2Fnotes%2Ftn28%2FUTN28-PlainTextMath-v3.1.pdf&data=05%7C01%7Cmurrays%40exchange.microsoft.com%7C853338add7794752ceaf08db0bbe6383%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C638116684718864473%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=7o9Kg3MOiSewBFKb1MUwa6JZC8VPQZ0VDm5MSCtfbtw%3D&reserved=0> more straightforward. I’ll need to add a “hidden field” to UnicodeMath to house intents in general, but that’s relatively clumsy compared to using symbols with the built-in semantics.

Thanks,
Murray

From: Neil Soiffer <soiffer@alum.mit.edu<mailto:soiffer@alum.mit.edu>>
Sent: Friday, February 10, 2023 1:13 PM
To: Deyan Ginev <deyan.ginev@gmail.com<mailto:deyan.ginev@gmail.com>>
Cc: www-math@w3.org<mailto:www-math@w3.org>
Subject: [EXTERNAL] Re: [www-math] <none>

> It would be quite informative for the group to understand why and where these inconsistencies arise, and if different implementers have strong reasons for their choices

This is easy to answer: because people have limited resources and they implement what they can and based on their guess as to what is most important and what they know. There are no rules or guides as to what to do and the space of options is vast. There are thousands of Unicode characters that might occur in STEM content. If you only have time to translate 200-300 of them, there is nothing out there at the moment that says "these are the most common ones" or even more specifically, "in calculus, these are the symbols that are used" (actually I do have a paper with some data on calculus textbooks, but it is only based on a few calculus books). Similarly, there is nothing out there saying "these are common notations". People usually implement squared, cubed, square root, and cube root, but commonality rapidly drops from there. There isn't anywhere that says "use ... from .. to  ... of ...' for large ops in an munderover, etc., so some systems might miss that.

As for inviting implementers...

I more or less represent what NVDA does since I did MathPlayer, and now MathCAT and those are what most people use with NVDA AFAIK -- there is also Access8Math addon but I don't think it is widely used based on people I've talked to.

The other screen readers are JAWS, VoiceOver, and ORCA (linux). I know a contact at JAWS I can ask about joining a call. Based on past experience where other W3C groups asked for someone to talk with them, there is only a small chance that will happen. But if the group is interested (I'll poll the group at the meeting next week), I'll ask.

I don't know anyone in the VoiceOver group. Apple does have a representative in the ARIA group -- maybe he can suggest someone. It's a long shot they will be willing to join for a call -- Apple tends to be very tight-lipped/standard committee avoidant.

I've been trying to find someone to talk to in the ORCA community about their math support. I know a previous developer, but she hasn't worked on math in years and has moved up the W3C food chain and is extremely over committed now and said she can't help me; again dubious she would want to talk to a group given she doesn't work on math accessibility anymore. AFAIK, no one has worked on math accessibility in ORCA in years.

In the group, both Steve and Sam have been in the AT community for years; Sam is an AT developer. So we do have some pretty good expertise in the group wrt to accessibility.

    Neil


On Fri, Feb 10, 2023 at 12:37 PM Deyan Ginev <deyan.ginev@gmail.com<mailto:deyan.ginev@gmail.com>> wrote:
Hi Neil,

Would it be possible to invite multiple AT representatives from the systems in question to the group? Maybe not as permanent members (as that is a large admin burden) but at least for a small invited talk each?

It would be quite informative for the group to understand why and where these inconsistencies arise, and if different implementers have strong reasons for their choices.
Some would consider that as a prerequisite to standardizing one outcome over another, which is what a "Default" ruleset represents.

It would also be helpful to get written testimonies of the affected AT users, so that we can get a clearer view of their pain points. Consider the way arXiv excerpted their user study as five "Themes" at the end of their accessibility report:
https://info.arxiv.org/about/accessibility_research_report.html<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Finfo.arxiv.org%2Fabout%2Faccessibility_research_report.html&data=05%7C01%7Cmurrays%40exchange.microsoft.com%7C853338add7794752ceaf08db0bbe6383%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C638116684718864473%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=xc67BzRo5g1lot04wXD3xdEt74ooHuz99TWyIe3MlM0%3D&reserved=0>

It is easy to believe there is room for improvement, but hard to see the constructive actions our group could take without more thorough preparation.

Greetings,
Deyan

On Fri, Feb 10, 2023 at 3:26 PM Neil Soiffer <soiffer@alum.mit.edu<mailto:soiffer@alum.mit.edu>> wrote:
I was part of a virtual STEM accessibility conference for the last two days. In the wrap up, a few people complained that there isn't a lot of consistency among AT. Some read one character ok (in MathML) and another won't read it. I'm not sure whether anyone complained about inconsistency with the speech other than dropping characters (JAWS seems to drop parens in many cases where it shouldn't) or speaking something ambiguously. I volunteered that our group was considering issuing some baseline guidance to AT as to minimal support they should have and people felt that was a good idea.

This relates back to defaults. Whether we add something to spec in an appendix or produce a note, it seems like the AT users at least feel it would be a good idea to have some minimal baseline all AT should support. I suspect that it would also be helpful for AT developers so they know "this is the important part" -- don't skip support for these notations and these characters.

    Neil