Claus Kick wrote:
> Cesar Rabak wrote:
>
> *snip*
>
>
>> Due this, it would be interesting to have a mechanism similar to the
>> Google one that searches for near 'words' and asks if the intended word
>> is not a listed in the choice menu.
>>
>
> slightly OT:
>
> Are there any "open algorithms" how word approximation is being done?
>
There are many. The Levenshtein Distance is the most common technique
used to identify spell corrections. Making it fast for large dictionary
sets is the tricky part. Levenshtein Automata are generally used, or you
can go a step further and use more CPU (but less memory) and do
Immitation Levenshtein Automata as is done in the Spellchecker2 package
in public store.
It's worth noting that building a dictionary of all the selectors in a
Smalltalk image would take a fair amount of time - and keeping it up to
date would also take time. The RB includes a "did you mean" fixer for
class names and shared variable names, etc - but its set is so small it
does it the naive slow way, which is fast enough considering how few of
them there are in the image.
Cheers,
Michael
_______________________________________________
vwnc mailing list
[hidden email]
http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
Locked
