Mozc

From ArchWiki
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

From the project's home page:

Mozc is a Japanese input method editor (IME) designed for multi-platform such as Android OS, Apple OS X, Chromium OS, GNU/Linux and Microsoft Windows. This OpenSource project originates from Google Japanese Input.

The differences between Mozc and Google Japanese Input are described in detail at the project's About Branding page, but in short, Mozc's open-source code does not include Google's extensive word conversion tables (so called "dictionaries"), so its conversion quality is not equivalent to that of Google's branded product. This can be mostly mitigated with custom dictionaries though (see below).

Installation

Install ibus-mozcAUR, fcitx5-mozc, fcitx-mozc or emacs-mozcAUR.

Split packages

As an IME, Mozc has two distinct parts: a server, which does all the work of the conversion, and a module, which allows the server to communicate with an input method framework like IBus, Fcitx5 or Emacs (which includes its own IMF) and output the result on the screen. A separate module is required for each IMF.

Some Mozc packages split the server from the modules and provide it as a separate mozcAUR package, allowing for different modules to be installed side by side or removed, independently from the server itself (this can be useful for example in a multi-user setup where different users use different IMFs).

Other Mozc packages forego this split and instead bundle the server together with a module. This does have its advantages, but one very notable disadvantage is that it makes it impossible to install additional modules, due to file conflicts between the packages.

If for some reason you will be needing more than one IMF module installed at the same time, the solution is to either build the non-split packages yourself and manually resolve the file conflicts, or to only choose from among the packages that split the server from the modules.

The UT Dictionary

As already mentioned, Mozc's conversion quality is not quite as good as that of Google Japanese Input because it doesn't include Google's extensive word conversion tables (so called "dictionaries"). A solution exists for that in the form of the UT dictionary, which is a third-party dictionary that enhances Mozc's conversion quality and brings it closer to Google Japanese Input. It achieves that by including thousands of additional words aggregated from several popular online sources (based on page rankings from Google, Yahoo and Wikipedia) and also by integrating a variety of other specialized sources such as the NEologd dictionary, which in the words of its creator contains "neologisms (new words), which are extracted from many language resources on the Web".

Tip: Use of the UT dictionary is optional but highly recommended.
Warning: In the past, there used to be a dictionary called UT2 which was supposed to become the successor of the UT dictionary and be used in parallel with the NEologd dictionary. As of November 2020, both UT2 and NEologd have instead been merged together with UT and are now provided as part of it by default. Due to this history, there still exist some packages in the AUR which are based on the old UT2 and/or the NEologd dictionaries; they can be recognized by their names containing 'ut2' or 'neologd', and they should be considered deprecated, unsupported and extremely outdated. When in doubt, the modern UT packages should always be preferred.


The following table shows the various available packages; purple cells indicate combined packages, while yellow cells indicate packages containing toggles for additional modules. Some of these packages can also be found in the pnsft-pur repository.

Fcitx5 Fcitx IBus Emacs
Official fcitx5-mozc fcitx-mozc ibus-mozcAUR emacs-mozcAUR
UT fcitx5-mozc-utAUR fcitx-mozc-utAUR ibus-mozc-utAUR emacs-mozc-utAUR
fcitx-mozc-ut-unifiedAUR ibus-mozc-ut-unitedAUR emacs-mozc-ut-unitedAUR
Note: Once Mozc is installed, you may need to restart X or your input method framework before you can use it.

Configuration

IBus

See also IBus for IBus configuration.

If you use Mozc by default, set it up via ibus-setup:

$ ibus-setup

Choose Input Method tab and move Mozc to top of the list.

You can switch input method by Alt+Shift_L (as per the IBus default).

Tip: When used with IBus, by default Mozc will launch in Direct (Eisu) input mode, which can be problematic on keyboards that lack the Eisu key. To change this behavior and make Mozc launch in Hiragana input mode instead, edit ~/.config/mozc/ibus_config.textproto and append the following line:
~/.config/mozc/ibus_config.textproto
...
}
active_on_launch: True

uim

See also Uim for uim configuration.

Configure uim preferences by running:

$ uim-pref-gtk (Or, uim-pref-gtk3/uim-pref-qt4)

which brings forth a GUI.

Choose your preferring input method as Default input method.

Note: Mozc will be not listed in Default input method at first time so you will need to add it into Enabled input methods to use.
Warning: You must run the following command whenever you upgrade or (re-)install uim.
# uim-module-manager --register mozc

Fcitx

See also Fcitx for Fcitx configuration.

Open the configuration dialog of Fcitx by running:

$ fcitx-configtool

In the Input Method tab, click on the plus sign and choose Mozc from the list in the dialog. Depending on your configuration, you might need to disable the Only Show Current Language option for Mozc to be available. After confirming the dialog, you can activate Mozc as input method using the usual keyboard shortcuts.

Mozc for Emacs

You can use mozc.el (mozc-mode) to input Japanese via LEIM (Library of Emacs Input Method). To use mozc-mode, write the following into your .emacs.d/init.el or some other file for Emacs customizing:

(require 'mozc)  ; or (load-file "/path/to/mozc.el")
(setq default-input-method "japanese-mozc")

mozc.el provides "overlay" mode in the styles of showing candidates (from mozc r77) which shows a candidate window in box style close to the point. If you want to use it by default, add the following:

(setq mozc-candidate-style 'overlay)

C-\ (toggle-input-method) enables and disables use of mozc-mode.

Disabling XIM on Emacs

When you are using input method on your desktop and assigning activation/deactivation of input method to C-SPC, you will be not able to use C-SPC/C-@ as set-mark-command on Emacs. To avoid this problem, add the following into your ~/.Xresources or ~/.Xdefaults. xim will be disabled on Emacs.

Emacs*UseXIM: false

Tips and tricks

Confirming Mozc version which you are using now

Type "ばーじょん" ("version") and convert it while activating Mozc. The version number of Mozc will be shown in the candidate list like follows:

ばーじょん
バージョン
ヴァージョン
ばーじょん
Mozc-1.6.1187.102  ⇐ Current version of Mozc
...

Launching Mozc tools from command line

The followings are commands to launch Mozc tools.

  • Mozc Settings: $ /usr/lib/mozc/mozc_tool --mode=config_dialog
  • Mozc Dictionary Tool: $ /usr/lib/mozc/mozc_tool --mode=dictionary_tool
  • Mozc Word Register: $ /usr/lib/mozc/mozc_tool --mode=word_register_dialog

Use CapsLock as Eisu_toggle key on ASCII layout keyboard

In all of the preset keymap styles of Mozc, the command Toggle alphanumeric mode on Composition mode is assigned to the Eisu (Eisu_toggle), Hiragana/Katakana or Muhenkan key, but the ASCII keyboard has none of them.

One solution for it is to use Caps Lock key as Eisu_toggle (Mozc does not recognize the Caps Lock key as of r124). The following is a way to assign Eisu_toggle to Caps Lock (without any modifier keys) and Caps_Lock to Shift+CapsLock, as on the OADG keyboard layout.

Warning: This will affect all applications.

Edit ~/.Xmodmap as follows:

keycode 66 = Eisu_toggle Caps_Lock
clear Lock

Then, restart X or run xmodmap to apply the changes immediately:

$ xmodmap ~/.Xmodmap

Troubleshooting

Building Mozc fails (process is killed)

If the build process fails with an error message like the following:

...
/bin/sh: line 1:  xxxx killed
...
make: *** [xxx/xxx...] error 137
...

Make sure you have not run out of memory.

New version of Mozc does not appear though I upgraded Mozc and restarted X or Input Method Framework (not rebooted)

The old version of Mozc may be still on your memory. Try to kill the existing mozc_server process:

$ killall mozc_server

mozc_server becomes defunct

Mozc cannot run in root. Start X in normal user.

mozc_emacs_helper not found in emacs

When installing mozc.el you need to install a helper program called mozc_emacs_helper.

You need to install emacs-mozcAUR or emacs-mozc-ut2AUR for this helper program.

mozc tool(s) does not launch

It is also possible to launch Mozc tools with log sent directly to console, which is useful for debugging. To enable this, simply append --logtostderr like this for example:

$ /usr/lib/mozc/mozc_tool --mode=config_dialog --logtostderr

See also