Re: Freedesktop accessibility draft

[Eric Johansson <esj eggo org>]

On 6/29/2019 4:02 PM, francisco del roio via gnome-accessibility-list wrote:
> Hello,
>
> I was reading this document[1] and I have to say something about TS[2]
> review section.

I have a few comments on the speech recognition section. I'm not sure
where/how to make them visible so I'm piggybacking them here.


I am glad to see the inclusion of the mark and point concept. Simply
getting people to understand why this is important has been a
long-standing frustration of mine. However the rest of the section needs
some significant enhancement.

A fundamental concept is that a speech user interface tends to be wide
and shallow where is a GUI tends to be deep and narrow. Another concept
is that speech interfaces don't control everything about an application.
It seems like it would be better to break speech interfaces into sub
applications as of disambiguate a grammar without making it too wordy.

The user should be able to construct their own grammar in order to
implement a task specific speech user interface. At a minimum there
should be open-loop grammars. By this I mean a grammar associated with a
specific application or window but has no other information to inform
the grammar. A better solution would be a closed-loop grammar based on
the state of the application and the ability to control the state of the
application so the speech interface can take a more direct route to
controlling the application. A close group grammar also has the
advantage of reducing cognitive and vocal load. What you see is what you
say.

The system needs to differentiate between dictation and commands. For
example if you are dictating, single character commands in the
application should be disabled. For example, if you dictate "the quick
brown fox" into Thunderbird, what happens to your email?

There are more things but this is all I can write up at the moment.