Datasets:
Correct Turkmen and Gagauz alphabets; set licence
Correct two alphabet errors and set a licence
grapheme_sets.py carried a spurious e-acute in the Turkmen alphabet, which
inflated |G_L| for tk/tr from five graphemes to six. The 1999 standard
(TDS 565) has 30 letters and no e-acute. Gagauz used U+021B (comma below)
where the standard letter is U+0163 (cedilla), matching the c-cedilla and
s-cedilla of the same orthography; these are distinct code points and a set
difference over the wrong one is wrong.
gl_pairs.csv and containment.csv regenerated from the corrected alphabets.
Only the tk/tr and gag/tr rows change: tk/tr goes from 6 graphemes to 5.
All alphabets are now marked verified, and the module docstring records the
four conventions that each change a published number if taken the other way:
promulgated alphabets over usage-based inventories, the 1998 Tajik decree,
Uzbek digraphs as sequences, and the Gagauz cedilla.
Licence set to CC BY 4.0, matching the paper.