The modes showed their raw IDs (com.autozhuyin.inputmethod.AutoZhuyin.Auto).
InfoPlist.strings (zh-Hant, en, Base) now name them 自動注音(中英自動) and
自動注音(英文); build.sh copies the .lproj folders into the app.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
macOS: the input method now has two input modes in Info.plist, 中英自動 ("注")
and 英文 ("A"), each with its own menu-bar icon. A Shift tap with nothing typed
switches the mode (TISSelectInputSource), and choosing a mode in the menu bar
switches the engine (setValue:forTag:). --register enables both modes.
Windows: the taskbar IME icon switches between icon.ico and icon-en.ico.
Linux: the IBus mode property also switches its icon.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
After an English word, 23 of 50 common first-tone characters (機, 吃, 書…)
were read as English when space was pressed: the fixed +1.5 context bonus
was far too strong for the trained weight scale, and the training data never
contained the "space just pressed on a first-tone syllable" decision.
Context is now the ctx.zh / ctx.lit features, trained with the rest; the
synthetic data includes the previous mode and the first-tone space decision.
Test accuracy: commit 98.22% -> 99.52%, typing 88.73% -> 91.49%, and when the
language differs from the previous word 92.0% -> 99.3%.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With "Latn" in tsInputMethodCharacterRepertoireKey macOS treated AutoZhuyin
as ASCII-capable and kept it active in secure text fields, so typed passwords
showed up as underlined composition text. Apple's Zhuyin and vChewing declare
only Hant (not ASCII-capable); macOS then switches to an ASCII keyboard layout
in password fields and the input method never sees the keystrokes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Scoring is now features x weights (Features.swift / features.py), with the
old hand-tuned values as defaults. training/ generates synthetic mixed
Chinese/English buffers from the lexicon and word lists and trains the
weights with an averaged structured perceptron over the same segmentation
DP the input method runs. On 10,000 test buffers commit-time accuracy goes
from 97.58% to 99.62%; typing-time (prefix) accuracy from 86.15% to 87.81%.
Both apps load Resources/classifier-weights.json. A parity test checks the
Swift segmentation matches the Python one on 400 fixtures.
Also fixes space after a toned syllable being scored as an extra stray tone,
which could turn short Chinese like "u4 " into English.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>