QazaqKey vs. the official Latin alphabet

The headline number

Across a 139,621-word corpus spanning classic literature, Soviet-era narrative, modern slang, legal text, and proper names, the official alphabet carries 3.2× the non-ASCII character load of QazaqKey — 50.2 diacritics per 100 words vs. 25.9, or 83.3 vs. 25.9 if the official alphabet's dotless ı is counted (it should be: dotless ı is what breaks casing, search, and default fonts). QazaqKey does this with exactly four non-ASCII letters in practice — ä ö ü ń — against the official alphabet's eight.

Full genre-by-genre analysis, including where the official alphabet does better (legal/proper-noun text, before QazaqKey's citation-mode fix — see below) and where QazaqKey has real, acknowledged rough edges, is in the project's full evaluation report. This page shows a representative sample, not the whole corpus.

Classic literary prose (Abai)

Native, diacritic-heavy text is where the gap is starkest — every additional letter the official alphabet needs is a letter QazaqKey doesn't.

Legal / official register

This used to be QazaqKey's worst genre — unassimilated legal and administrative loanwords were run through native vowel-harmony rules that don't apply to them, producing spellings like Konstuwcyja. A dedicated citation-mode rule now catches these, matching the official alphabet's handling of the word itself while keeping QazaqKey's much lighter diacritic load everywhere else in the sentence.

Foreign proper names

Country names converge with the official alphabet once assimilation-depth rules apply — but native Kazakh names (last two rows) stay in QazaqKey's cleaner default register rather than picking up the official alphabet's dotless ı.