Data behind the layout
212,000,389Letters and clusters counted
20,646,826Coeng clusters
142Units ranked
9,428Words for letter pairs
3,000Practice words
0.9996Ranking stability
Modelled metric
Bigram figures currently use word-list statistics as an interim estimate. Re-running the optimiser on corpus bigrams may change the layout.
Most frequent keys in the selected layout
How it was made
1Count real text
About 212 million Khmer letters and coeng clusters were counted from a text corpus. Split-half checks show the ranking is stable from roughly one million letters.
2Model the effort
Each key has a reach cost. Shift and AltGr add a penalty, and letter pairs are scored for same-finger repeats, rolls and hand alternation. Fingers have load caps.
3Search for the best fit
Simulated annealing tries over a million key swaps. Dedicated cluster keys like ្រ did not lower effort with only 33 base keys, so the default layout types them as coeng + consonant.
4Make it learnable, then measure
Paired letters such as ក / គ sit on the same or neighbouring keys. Export your keystroke log to compare layouts with real typists.
Experimental layout, not an official Khmer standard.