Re: Spell checker optimization
>You mentioned at the beginning about the ACTUAL comparison routine, to
>pick the best word.
This thread has gotten a little big.... And without quoting... Where
exactly are you referring to? (I've finished the comparison and
replacement word routines, I just need to speedup the dictionary load and
perhaps make the dictionary filesize smaller)
>I wrote a QBASIC fuzzy search routine a while ago which would do just
>the job.
>It is already pretty well optimised for typical mis-spellings (letters
>in wrong order like hlelo or missed out letters) and a little extra code
>to check for keyboard typos (s = a or d, k = j or l)
Mine does this, but a little more generic. (With the detail David Cuny
went into on the message he sent me, I figure he could use the
information to create a spell checker himself if he had more time....) It
checks for missed out letters and wrong order, as well as extra letters,
single letter typos and words that should be two words. It could be much
more optimized, but like I said, on the 486 it's response time for the
check is 'instant'. I don't notice any delay. (Except that 16 second one
while the 1/2 meg dictionary is loading....)
>I'll convert it using David Cuny's ebasic, and then tidy up if you
>think it might be useful. I never bothered to turn it into a spell
checker
>because QB was SO SLOW - I mean I even had to write my own sort
rountine...
The replacement routine I have now works great. Later on I might add
soundex encoding to replace words that are too mangled for the current
checks. (Mostly big words that sound the same to soundex but have too
many messed up letters, but I've only had 2 words so far that I didn't
get the correct replacement in the list)
>own sort rountine...
^^^^^^^^
My spell checker would catch that and give the correct replacement,
assuming of course, that "routine" is in the dictionary... :)
_____________________________________________________________________
You don't need to buy Internet access to use free Internet e-mail.
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]
|
Not Categorized, Please Help
|
|
