Re: [tlug] Japanese input method

[Travis Cardwell <travis.cardwell@example.com>]

On 2015年07月11日 00:13, Stephen J. Turnbull wrote:
> Travis Cardwell writes:
> 
>  > I have a full plate, but this gives me an idea...  Somebody should do a
>  > presentation about the development side of [Japanese] input methods, of
>  > which there a number of interesting aspects.
> 
> Specifically?

One topic that I would like to learn about is how input managers interface
with the UI.  What protocols, if any, do X and the various GUI toolkits
(GTK, QT, etc.) provide?  How is composition and candidate selection
rendered?  What are the challenges of making a decent, cross-WM toolbar
widget? [1]

Another interesting topic is "learning."  What techniques are currently
used?  What does the proprietary software have/do that makes it better
than the open-source implementations?  What are possibilities for
improvement?  One feature that I really like is the ability to easily
"unlearn" specific things. [2]

Travis

[1] I have been using UIM for years, and the toolbar widget has never
displayed correctly in XMonad or i3.

[2] I sometimes have to hack my ~/.anthy/last-record* files in order to
remove mistakes.  For example, one time I somehow got Anthy to "learn"
some seemingly-random string of ひらがな and カタカナ when I wanted to
enter 「株式」, and I was subsequently unable to enter 「株式」 (in that
it was no longer a candidate).  Until I had a chance to hack the Anthy
files, I simply entered 「株式会社」 and removed the last two characters.

I am not sure how that happened, but such issues are not uncommon.  My
work computer currently has such an issue.  (I have to switch to カタカナ
mode in order to input 「スクリプト」, which is annoyingly a word that I
use relatively often at work.)  Since the Anthy files are locked while
using the program, I have to wait for a convenient time to modify them,
and, in the meantime, find alternate ways to input problematic words.
When Anthy "learns" mistakes, I then have to "learn" how to input around
them. :|