Re: Vision impaired interface

From: Matt Shomphe <mshomphe_at_comcast_dot_net>
Date: Thu Oct 30 2003 - 23:10:51 CST

At 09:06 PM 10/30/2003 -0800, you wrote:
>At 8:25 PM -0800 10/30/03, Matt Shomphe wrote:
>>At 10:12 PM 10/30/2003 -0600, you wrote:
>>> > though. For example, the first that came to my mind was, how to
>>>> handle write-ins - any ideas?
>>>
>>>Write-ins present a number of different and difficult problems.
>>>
>>>First, for voters who are good at typing. There are many examples of typing
>>>interfaces for blind users. These interfaces for instance allow the user to
>>>choose to type with no feedback, with line by line feedback, with word by
>>>word feedback, or with key by key feedback. For any particular interface,
>>>it takes a while for the users to learn the commands, but this investment
>>>has a long term payoff. Users typically turn off most of the feedback once
>>>they get good at the system. The voting system must be simpler, since it
>>>should not require a long training period. A suggestion would be to give
>>>all three - letter-by-letter, word-by-word, and end of line audible feedback
>>>with no options . Thus if the user typed in John Smith [and then hit
>>>ENTER], the computer would say: J O H N John S M I T H Smith [pause] John
>>>Smith.
>>>
>>>Then there would be an option to confirm or to go back and retry. All this
>>>is a straightforward software coding task
>>>
>>>For the user who is not a decent typist or can't type at all it is harder.
>>>One approach would be to use the same system, but to substitute letter by
>>>letter voice recognition. The user says "J" - the computer says "J - is
>>>that right" - the users says "YES" or "NO" - etc. Are there any General
>>>Public License voice recognition programs that could recognize the spoken
>>>names of letters - either speaker-independent or one where the user would
>>>have to say the alphabet for training?
>>>
>>>Word by word recognition would be too hard, since the writein name could be
>>>any arbitrary name. Many word recognition systems work by comparing words
>>>spoken to expected words. But this might result in "Sharon Egger" being
>>>transcribed as "Schwarzenegger."
>>>
>>>Peter Maggs
>>>p-maggs@uiuc.edu
>>
>>As the resident speech-guy, I guess I should throw my hat into this:
>>letter recognition is really really hard. (For David, "wicked friggin'
>>hahd"). Basically the "ee" set: "e", "b", "c", "d".... they all have the
>>same vowel which makes disambiguation hard; any freely available system
>>will not be able to handle that. And name recognition is amazingly
>>difficult, especially without visual confirmation. Furthermore, if we're
>>doing a speech interface, that kind of does away with the privacy of the
>>user, correct?
>>
>>What do blind people do now? Can we just revert them to an existing
>>system? This is a bit of an edge condition, non?
>>
>>M@
>
>You can respond, "b as in bat, c as in charlie, d as in dog" or whatever
>the preferred letter-words are these days. Do you mean "b as in bat or p
>as in pencil"?

Well, if you wanted, we could do the military alphabet. You'd have to
confirm after each letter. The search space for a word of n letters is
26^n (maybe 27^n if we include "space").

As I said, it's not a huge part of the population. Can they just use
alternative methods? Can someone else fill out the ballot?

Matt
==================================================================
= The content of this message, with the exception of any external
= quotations under fair use, are released to the Public Domain
==================================================================
Received on Fri Oct 31 23:17:06 2003

This archive was generated by hypermail 2.1.8 : Fri Oct 31 2003 - 23:17:07 CST