Carpalx optimizes keyboard layouts to create ones that require less effort and significantly reduced carpal strain!
Have ideas? Tell me.
I evaluate the desirability of a candidate keyboard layouts by considering the typing effort, along with the following characteristics. Layout characteristics are derived using the same corpus used to optimize it, or in the case of pre-existing layouts (e.g. Dvorak), using the English training corpus.
A keyboard layout associated with a lower typing effort is better (as far as measured by the typing effort model) than one with a higher effort. The layout is evaluatated independently of the model using figures of merits (FOMs) that fall into two groups: relative usage frequencies and runs. These FOMs are important because I cannot be certain that my typing model, or the parameters I have selected, will produce a desirable layout. By comparing the FOM values for optimized layouts against existing, well-established layouts, I can test my model.
Realtive usage frequencies describe how frequently a particular row, hand, or finger are used.
Runs describe how frequently the same row, hand or finger are used N times in a row. The cumulative distribution of runs is best used for this FOM. Given a value, n, the cumulative distribution of a run (e.g. left hand run) gives the relative number of hand runs shorter or equal in length to n. Row and hand runs are futher categorized by top, home and bottom (for rows) and left and right (for hands). Finger runs are not expressed for individual fingers (e.g. a finger run of 5 indicates one of the fingers was used 5 times in a row). The row jump FOM is the total number of rows traversed by typed fingers within a hand run. The row jump distinguishes the case of typing ttt (hand run = 3, finger run = 3, row jump = 0) and tbt (hand run = 3, finger run = 3, row jump = 4).
Desirable layouts have most of their runs very short, with the exception of the home row runs (rrh(n), i.e. the number of times keys of the home row are hit n times in succession), which should be lengthy.
Detailed typing statistics, such as those described above, unambiguously characterize a layout's performance. However, these values can be hard to interpret and translate to typing in practise.
To give a more subjective impression of a layout's fitness, I have ranked words based on the effort (per unit length) required to type them. I have ranked words for QWERTY, Dvorak, Colemak, QGMLWY (fully optimized, Colemak-like), TNWMLC (worst layout for English).