[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Speechd-el: Word Echo, Smarter Navigation, Visual Performance?

From: Veli-Pekka Tätilä
Subject: Speechd-el: Word Echo, Smarter Navigation, Visual Performance?
Date: Sun, 10 Aug 2008 17:50:03 +0300

Hi, my first post here:
I've run into three speechd-el problems that are rather major in my view, 
and was wondering if anything could be done about these.  In brief they are, 
lack of word echo, lack of smart reading based on navigated unit, and visual 
performance on older machines. Regexp search for ^\d+ (PCRE-syntax)  if you 
want to jump to the next point.

1, lack of word echo: There's no word echo in speechd-el for immediate 
feedback when typing,  That's just majorly bad for me, since I've edited 
with word echo about 10 years in quite a number of apps. Even with the 
fastest Espeak setting, char echo is too slow for me and it is hard to 
mentally parse the individual letters to words. I can pick out most ENglish 
and Finnish typos by pronunciation when word echoed, but character echo 
doesn't make that process quite as straight forward and transparent.

2. Smarter navigation reading: consider moving in sentences. Most editors 
don't let you to. Reader commands for the same thing, if any, are reader 
specific. However, when grasping difficult text where you need thinking 
time, or for snipping stuff, moving in sentences is also advantageous. 
Normally you would have to read the whole paragraph (physical line), or 
being forced to pause based on arbitrary logical line breaks of a file 
dictated by the screen size. With smart reading, you could take as long 
pauses as needed, and tell Emacs to read the next desired unit back to you 
when ready.

SO ideally I'd like to tell Emacs, move to the next sentence, and have 
speechd-el read me the whole of the next sentence, as that's what my 
movement implied. It does not do that, reading the destination line after 
the navigation in stead.

This same thing could be useful for larger units. Navigating a paragraph 
would read the paragraph, navigating a chapter would read the chapter, with 
no need to learn any reader specific keystrokes. And if you only wanted to 
hear, say the heading and not the content, you could turn this off or 
interrupt the speech by another navigation command.

This reading style requires that the reader understands the app logic of 
what unit you've moved to and can ask the app for the next/prev such unit. 
But I thought speechd-el could easily ask Emacs for the units, so surely 
that's mnot the problem.

Alternatively, on the reader side, you could just read whatever is between 
the current and previous cursor position when the cursor moved, which would 
amount to much the same thing without the reader having to understand app 
logic other than getting the whole doc contents. There are some problematic 
boundary cases such as first and last unit, but still I think the genral 
idea might be useful and so far unimplemented.

One nice addition would be a command, stop the speech and drop the cursor to 
the pos the reader was reading at the moment. It would be pretty neat in 
hands-free reading a whole chapter, wishing to pause in the middle for 
slight editing. Later on, you could then continue from where you left off by 
using the next section command again.

3. Performance on older machines: When I touch type a sentence fast in 
speechd-el the result is quite curious. FIrst it is as though nothing 
happened, then after about three sec all text is appended to the line, the 
visual cursor jumps ahead quickly, and very little is spoken. Can anyone 
replicate, how is the behavior on newer hardware? It would be nice if the 
visual cursor was snappy, too, since I'd like to use real magnification or 
the GNome terminal with very large font sizes. I use speech primarily with 
some magnification added - 7x full screen in Orca.

So far, by the way, I've yet to find terminal magnification. Orca doesn't 
know how to trakc things in Emacs. The Gnome-terminal refuses to use large 
font sizes such that it ads scrollbars keeping the console dimensions: say 
80x25 at 72 pt. Finally, other magnification I've heard of would require the 
real text-moed console as opposed to running in Gnome. But that's slightly 
OT anyway.

On an introductory side note, I'm a long time Windows power-user and Perl 
progremmer. I do hope Emacs with speechd-el would  be the Linux killer app 
for me, a real good text editor. I'm already a huge regexp fan in editors, 
and very keen on the structural navigation that Emacs offers sentences, 
paragraphs, units that depend on document type, and so on.

I'm running Ubuntu hardy with the version of Speechd-el coming with it after 
the latest updates, on a Pentium M laptop with a gig of RAM. I'm using 
Espeak in speech-dispatcher with the fastest speech rate.  I'm very new to 
Emacs, however, and my LInux knowledge is restricted to a bit of POSIX 
programming in C, as well as what the book Learning Unix teaches one. I also 
know Perl and LaTeX, but these aren't really Nix specific.

Hope I'm making this clear.

With kind regards Veli-Pekka T?til?
Accessibility, Apps and Coding plus Synths and Music:

reply via email to

[Prev in Thread] Current Thread [Next in Thread]