Re: [g-a-devel] GOK and a different way of show keyboard

[Bill Haneman <Bill Haneman Sun COM>]

Hi Rangzen,

It's very nice to hear from you, and thanks for your work on pylisiere.

I agree that we should try to work together.  There are many problems 
which onscreen keyboard will have in common, and GOK is already well 
developed.  Perhaps you saw the '1.0' announcement that was sent to this 
list very recently.

While it is true that scanning a full 'compose'/'qwerty' keyboard one 
key at a time is too slow, GOK supports a variety of scanning methods.  
The row/column scanning which is currently offered for single-switch 
users is much faster, and GOK uses word completion (including 
user-specified dictionaries, and 'learning'), but the maximum input 
speed for this method is still only about 8 words per minute or so. 

One thing which you've probably already discovered is that GOK supports 
multiple 'access methods' for activating its keys.  I've mentioned one, 
above ('automatic scanning'), but for some users with 'locked in 
syndrome' there may be other choices, for instance there have been 
previous experiments connecting a device activated by breathing to a 1-d 
valuator, which demonstrate that useful scanning algorithms can be 
constructed with only one axis, with or without an additional switch.  
The main point is that different 'access methods' will suit users with 
different needs.  GOK is designed to allow new access methods to plug in 
to its framework, and this is an area where you could definitely help us.

While it would certainly be possible to present a set of 
multiple-character buttons using GOK, to produce something that looks 
similar to pylisiere, it might make just as much sense to adapt a new 
access method to preserve the 'qwerty' style keyboard but select the 
keys in something other than a row/column order.  It would also be quite 
straightforward to display a cellphone-like keypad with multiple letters 
and use a 'T9' technique to fill GOK's word prediction row.

In other words, there are multiple ways in which the work you have done 
with pylisiere could be combined with GOK, and I think it's a great idea 
to do so.  I'll be happy to help you in understanding the GOK source 
code, and I'm sure David feels the same way.  A 'T9'-like approach is 
something we've discussed before, and in particular if you can provide 
us with the T9 prediction engine I think it can be integrated fairly 
quickly.

best regards,

Bill

rangzen wrote:

> Hello to all the list,
>
> first of all : thank you for your work !
>
> and second : sorry for my english :)
>
> I work on pylisiere https://gna.org/projects/pylisiere a virtual keyboard.
> When i start this program my first goal was to find a way to write for
> people who can only use a binary signal to communicate like people under
> Lock-in syndrom or who can only move a finger (up or down, mouvement or
> not) or control their breath (breath or not, in or out).
> For this, a virtual keyboard wich look like our keyboard with more than
> 100 keys is unusable, even with automatic advance in the key, it's too
> slow ...
>
> My works now looks like the ditionnary who help you to write on cell phone
> with predictive words and more than 1 letter per key. This "technlogy" is
> called T9 in cell phone. I show my works to an hospital on a livecd wich
> is very interested and propose me some tests with disabled people to test
> their use of the program.
>
> There is a group of letter by touch so you have to use a dictionnary to
> propose the words wich can match with.
> Example :
> Key 1 : eairs
> Key 2 : tnoc
> You press 1,1 and 2
> Wich word match with [eairs][eairs][tnoc] : set, sin, etc order by number
> of use or freq of use in your langage, etc.
> You can see a screenshot at http://home.gna.org/pylisiere/
>
> Pylisiere can be change to work with any group of key per boutton, with
> any number of button. My work is actually on building free (as freedom ;)
> ) dictionnary.
>
> pylisiere can only input text and the hospital tell me that people want go
> on internet, watch dvd, etc. So i continue my research and i found gok ...
> You have probably more than 2 years of advance on me :)
>
> I think that it's probably better to adapt my job to yours, so i want to
> know if you have allraedy some works in progress for a vk with group of
> letter and dictionnary ?
> Is it possible to adapt this idea to gok ?
> Did you have already try this ?
> Is it too big to change ?
>
> Thank you very much.
>
>