* Using the letter frequency data from the:

* Using the letter frequency data from the:
  "UCSC LTRL Sinhala Corpus Version 1 June 2006�� (7M words).
  The total number of letters is 33880970. All percentages are
  rounded to 2dps.

* Assign the most frequent letter within a grouping to the unshifted
level,
  then the next most frequent to the shifted level, then the alt-gr
level,
  and the least frequent to shift plus alt-gr level. However, if
Alpapraana
  and Mahaapraana pairs exist, keep them consecutive.

* Key 'd':
  (U+0daf)      1006218 (2.97%)
  (U+0db0)      164704 (0.49%)
  (U+0da9)      171139 (0.51%)
  (U+0daa)      1229 (0.00%)

* Key 's':
  (U+0dc3)      1259498 (3.72%)
  (U+0dc1)      181189 (0.53%)
  (U+0dc2)      118730 (0.35%)

* Key 't':
  (U+0dad)      1568636 (4.63%)
  (U+0dae)      68688 (0.20%)
  (U+0da7)      712353 (2.10%)
  (U+0da8)      11415 (0.03%)

* Key 'x':
  (U+0db3)      120051 (0.35%)
  (U+0dac)      10360 (0.03%)