Lobste.rs user minimax gave me the tip of two input systems that provided me with a lot of interesting input. Edgewrite is an alphabet designed for input via stylus, optimized for this particular area. The other one being hexinput. More on hexinput later on, but I suspect it to be a method I saw referenced in a Dasher paper but couldn’t track down myself.
While on the subject of lobste.rs forum I must say I am so impressed with that forum, each post I do there gives back so much! Truly a high quality forum! I’ll do a post where I compile all the ideas I’ve received from you guys and any contributors will have a mention in my book.
So in short the alphabet is constructed by taking the roman element and presenting it as series of strokes that can be done through four edges. This in order to beat Fitt’s law as a movement towards an edge makes overshoot a non issue.
Perhaps more interesting is the concept of Universal Design that originated within architecture, stating that disability is a spectrum. Changes made to make buildings say wheelchair friendly also would make them easier to move stuff into and out of. Makes a lot of sense and applied to UI it means that a design that can be used by someone with a motor dysfunction will help out a “normal” user while situationally disabled, for example while walking.
So more on this on my draft for my chapter on writing recognition:
Chapter IV: Writing recognition
Like what I am doing?
Please consider supporting me, by Patreon, by sharing or by commenting!
My own work
As you’d might expect, I have some ideas of my own regarding text input. Those culminated in the sigma project released for the Christmas of ’18. Whereas I’ve failed to gather data in a meaningful scale, I’ve still come to some interesting conclusions on a qualitative basis.
In short, we’re talking about tree based structures that sort symbols by frequency and as such we achieve something akin to Huffman coding while typing.
One application that seems really interesting here would be editing LaTeX code. Here we see a niece tradeoff between power and usability I think. Try it out for youself
Freeing up the right hand for mouse-based editing makes a lot of sense here. The “paths” that are used often are quickly memorized and those forgotten are quite easy to find.
Other things that made a whole lot of sense was highlighting symbols by the probability predicted by the prediction engine. I ran a test on classical Chinese with a quad layout, where I managed to select from a huge character set at over 20 wpm using this system. So this might have some practical applications, not only for text but also for other applications that deals with huge data sets.
More in Chapter VI: Sigma project
I have some ideas on how to make a sigma-dasher-chorded hybrid that I think really could be something, so please stay tuned.
In other news
Also thanks to published author Cecilia Brunzell for helping me iron out some of the worst errors in my text.
In the cleanup branch at my github you’ll find a prediction engine that can predict writing on a character level. More interestingly it can handle multiple dictionaries and switches between them dynamically by weighting dictionaries according to recent accuracy. Build dictionaries by running make in corpus/wikisample.html and open web/boot.html.
Want to help bringing forth the keyboard of tomorrow quicker? Comment if you have ideas, share to get more people involved and or donate if you can.