Xavi Ivars <[email protected]> writes:

> 2013/2/21 Per Tunedal
> <[email protected]>
>     I'm in favor of all means to faciliate contributions from laymen.
>     Choosing a proper paradigm is a difficult task and thus all help
>     is wellcome. Your proposal is very interesting as it let's the
>     computer do a great deal of the work.
>      
>     Just a thought: as a user, you cannot know what words are already
>     present in the dictionnary. Let the user have a guess, i.e.
>     suggest a word, and if it's not yet in the monodix, let him/her
>     have an other guess until there's a hit (or he/she has got enough
>     of guessing).
>          
>
>
> That's a good idea. I'll add that too.

If the user first types in the new word, the tool could give a list of
guesses based on how that word ends. E.g. if the dictionary has

box box__n
fox box__n
ox ox__n
fix fix__v

and the user types in "paradox", it could give suggestions sorted by
longest common ending of a word using that pardef (and sub-sorted by
frequency of pardef usage):

box/fox
ox
fix

Here two have the same ending, but one is more frequently used than the
other. Then, as you show the suggestions, you could apply them to the
input word:

user> paradox
computerbrain> 1. box, fox, …: paradox/<n><f><sg> paradoxes/<n><f><pl>
computerbrain> 2. ox, …: paradox/<n><f><sg> paradoxen/<n><f><pl>
computerbrain> 3. fix, …: paradox/<vblex><imp> paradoxed/<vblex><past> 
paradoxes/<vblex><pres><3sg>
computerbrain> Select 1, 2 or 3, or type a word to search for
user> 1

You could use a trie to quickly look up reverse lemmas to find matching
pardefs. But maybe this is taking it a bit far :-) A list of pardefs
sorted by frequency would probably be good enough for most usage.


-Kevin


------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_feb
_______________________________________________
Apertium-stuff mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/apertium-stuff

Reply via email to