[log in to unmask] wrote:
> About word prediction, word lists etc: As I wrote a while ago, I am a 
> bit amazed that there seems to be so little available of these and 
> related tools for support for people with writing and reading 
> problems. There is a range of quite established tools and techniques 
> on the Windows and Mac platforms that seem to be largely lacking in 
> the FLOSS-Linux environment. This is a bit strange, as this concerns 
> the largest area of special needs in education, and it must be a 
> problem when trying to promote FLOSS learning platforms for schools.

Educational tools are generally quite weak in the FOSS world. This is 
partly due to the lack of good free content. Most FOSS people are still 
programmers. They can write functional software, but are not good at 
adding the content that is needed in education. A word list for word 
prediction is a related case to this. That said, I think Dasher has a 
fairly good word prediction system.

> Ideally these tools should be developed as modules that can be used 
> either as add-on tools for direct use with standard apps - for 
> keyboard users, or hooked in as add-on functionality in on-screen 
> keybords etc. This is the approached we used for SAW, cooperating with 
> developers of pred. and wordlist programs to make these capable of 
> being hooked up to SAW as an alternative UI - e.g. Prophet, WordAid, 
> and Penfriend. I think this is an approach that should be considered 
> for SOK. I'll try to enter this on the wish list when I'm back at work 
> in a weeks time or so.

Either a plug-in or even a separate application that can work with SOK. 
Some people might want to use word prediction while using a physical 
keyboard or other input methods.

I would like to see a prediction system that integrates with your text 
editor (or web-browser or whatever) so that you can enter text fluently 
without disturbing pop-up windows. You could then go back and check the 
result before saving/sending in the same way you currently check the 
underlined words for spelling in many applications. The prediction 
program would make a best guess based on the letters you entered but 
also the grammatical context. You would check it at the end and the 
prediction would improve next time based on your corrections.  Similar 
to the way Dragon Naturally Speaking works.

I think this would be more efficient in use than current prediction 
systems with pop-ups. The delayed nature of the system also allows for 
two-sided predictions (looking at the words before AND AFTER the word in 
question), which should provide more powerful prediction (but a more 
complex algorithm).

- Henrik