Wolf Vollprecht <w.vollprecht(a)gmail.com> さんはかきました:
I have recently created a standalone version of the Open Source version of
the keyboard input prediction/correction found in Android.
Some things make it IMO better than hunspell for typing correction: Namely,
it takes into account the geometric close-ness of keys on the keyboard and
it also supports n-grams (contextual corrections) in the dictionary.
ibus-typing-booster already supports trigrams for a long time.
The hunspell dictionaries are only used as a sort of fallback
for input where no n-gram data is available yet.
If ibus-typing-booster is trained for a while with user input
(or a file with typical user input is read using the setup tool),
then the hunspell dictionaries are not really used anymore,
mostly the n-grams from the user input are used then.
The close-ness of keys on the keyboard is not used yet, this might
be an interesting addition.
I would be more than happy to see that in the ibus typing booster as
replacement or alternative to hunspell.
A prototype of the thing that uses DBUS to communicate can be found here:
Let me know what you think!
ibus-typing-booster mailing list
Mike FABIAN <mfabian(a)redhat.com>
☏ Office: +49-69-365051027, internal 8875027