Re: dim: debian input method. call for help ;-)
My mail system seems broken...
ha shao <firstname.lastname@example.org> writes:
> It looks like you just want to save everything a user has input into
> a database. When there is a new input, search the database for
> matchese. The bigram is used to index the database for faster
> search. Maybe add some frequency information.
Simpler things first. I really need something similiar to MS PinYin
input method under Debian GNU/Linux. A scratch, ya know. ;-)
> It helps but I don't think it will extrapolate things that a user
> did not input before.
The learning procedure could learn from articles, just as well as from
user inputs. If the user wanna initiate the database for some specific
tasks, like scientific inputing, s/he could feed up DIM a scientific
article first. Or when the user wanna write an erotic story, s/he
could first feed up DIM with an erotic story. IMHO, this is a bit of
better than a fixed database. What's your opinion?
> Is this too much simplification from your scheme?
Simpler is better. But really it isn't in this case. ;-)
BTW, I've updated the design docs for guessing functions a bit. You
could read it from CVS. ;-)
The project page is at: http://dim.sourceforge.net