uzmorph 0.9.0

First public release of uzmorph: an improved version of the Apertium Uzbek transducer (apertium-uzb, upstream commit 00522e3), packaged with a Python layer. It adds a confident guesser, a proper-noun layer and a corpus-trained reading ranker, and outputs UD-style lemma, UPOS and features. Input can be Latin, Cyrillic or the 2026 Latin alphabet.

Headline numbers (details and caveats in the README):

	uzmorph 0.9	upstream 00522e3 + CG	ZirhUzMorph 0.1.1
Top-1 lemma+feats, held-out test (1,731 tokens)	93.5	79.7	87.8
Dev split (1,062 tokens)	94.1	78.1	87.5
Token coverage, dictionary / + confident guesser	90.69 / 94.80	87.51	n/a

• The evaluation gold is mostly model-built, with a partial native-speaker check (see README, "Limitations").
• The evaluation sentences are not distributed.
• Non-word acceptance: 1% (dictionary), 17% (with the confident guesser).
• Generation: 0 of 462 checked verb forms give a spurious -ish- variant.

Install: sudo apt install lttoolbox, then pip install uzmorph-0.9.0-py3-none-any.whl. The sdist uzmorph-0.9.0.tar.gz is the full Corresponding Source of the bundled transducers (GPL-3.0-only).