Thanks, folks, all of this is very helpful. It sounds like we'll definitely just want to come up with a really simple scale (e.g., 0-100, 0-1), and then test it everywhere to make sure it's "close enough." Andreas Stefik, Ph.D. Assistant Professor Department of Computer Science Southern Illinois University Edwardsville On Thu, Dec 10, 2009 at 7:34 AM, qubit <lauraeaves@xxxxxxxxx> wrote: > Jaws 11 uses 2 numbers -- the first is some inexplicable number that is > probably the one used by the synth. The other is percent -- of what I'm not > sure -- perhaps the the percent of distance between the lowest and highest > speed setting for that synth. Mine currently says "112, 65%". Depending on > what I'm reading I can push it up to 75 or even 80 or more for familiar > text, but I usually leave it at 65. > Now really, does anyone listen to speeds over 80 for general reading? I > have heard claims. If they can, perhaps my hearing aids are getting in the > way, or perhaps they are not comprehending what they are reading...*smile* > I aim for full or near full comprehension. > Happy listening. > --le > > > ----- Original Message ----- > *From:* Chris Hofstader <cdh@xxxxxxxxxxxxx> > *To:* programmingblind@xxxxxxxxxxxxx > *Sent:* Thursday, December 10, 2009 7:12 AM > *Subject:* Re: A question on Screen Reader Speed Standards > > JAWS uses the actual speech rate setting as provided by the synthesizer. > So, for Eloquence, it's a value that is only meaningful to it and your > ears. Other synths use WPM or some other fairly arbitrary system like > Eloquence. > > I can't recall which year it was but we did a JAWS release that used a new > Eloquence wpm feature. The setting was miserably flawed and the EtI guys > pulled it out and we all returned to their internal scale. > > WPM is really hard to get right for a variety of reasons, especially when a > screen reader sometimes sends as little as a single syllable and other times > a line of text or a full sentence. If the synth was used to read longer > pieces with a more predictable pattern, wpm would probably be a bit easier > but as screen reader users need to hear every different typeof chunk of > information, prediction becomes very hard. > > I've noticed that the "Alex" voice on Macintosh does some fairly clever > word prediction to create an illusion of being faster than it really is. > It's a nice voice but when it gets the word prediction wrong it can be very > confusing as one doesn't know if he misspelled, mistyped or simply used the > wrong word. It doesn't make these mistakes often but I wish I could turn > the feature off while I'm editing and back on for reading long documents. > > cdh > On Dec 9, 2009, at 4:37 PM, The Elf wrote: > > Andreas, > > I have no clue what jaws rates its speech rate on, especially since it > changes as you use the radio button to change it, flipping between two > slightly different numbers that make no sense to me what so ever. > > my thoughts, yes make it a standardized method, and use the more universal > WPM rating, at least that one makes sense! > > good luck, > elf > proprietor, The Grab Bag, > for blind computer users and programmers > http://grabbag.alacorncomputer.com > Owner: Alacorn Computer Enterprises > "own the might and majesty of a Alacorn!" > www.alacorncomputer.com > > ----- Original Message ----- > *From:* Andreas Stefik <stefika@xxxxxxxxx> > *To:* programmingblind@xxxxxxxxxxxxx > *Sent:* Wednesday, December 09, 2009 9:26 AM > *Subject:* A question on Screen Reader Speed Standards > > Hi folks, > > I'm working right now on trying to build up our cross platform speech > engines for the Sappy project and am trying, specifically, to get NetBeans > to store custom preferences related to screen reader speed. On Mac, we > basically just pass a flag to the TTS engine with a number, which, I > suspect, is words per minute, although I'm not completely sure. On PC, > things appear to be quite different and I'm not sure about all of the open > source, and other, solutions out there (insert your favorite technology > here). > > My question is, what would people suggest for standardizing the numbers for > speed of reading we use for screen readers across all platforms? For > example, does each screen reader everywhere measure speed in a different > way? Should just put everything in words per minute and not worry about it, > translating any screen reader that doesn't comply through some kind of > calculation (if possible?)? Should we just standardize through some > arbitrary metric, like 0 is the slowest and 1 is the fastest, then test > everywhere to make sure those settings are "reasonable" and that the user's > system preferences are not disturbed? > > To be clear, remember that our tool has to, ultimately, be compatible with > every kind of screen reader, and should still work for the blind even if no > screen reader is present (or if the screen reader doesn't work well at all). > That's why I am asking, > > Thoughts are welcome, > > Stefik > > >