Re: Robert...EDS - questons/comments
Darn it, Kat, you went and made my brain work (such as it is) this
morning.
Many many years ago, when I was taking some Operations Research classes,
we talked about search methods, especially search that may or may not
have an end point. I don't remember much unfortunately.
You probably know more about this than I do, and if not, *somebody* here
does, since I'm sure it's in some 200 level undergrad course somewhere.
Anyway, one thing we talked about is narrowing the scope of the search
with assumptions. For example, if the word starts with a consonant as
spelled, assume it really starts with a consonant. Secondly, assume that
beginning consonants are phonetic clones or adjacent on the keyboard.
Right there you've slashed your search quite a bit.
Another option might be to use heuristics. Replace letter at random from
a set of rules like those above, compare it to a dictionary of real
words, and if it fits, go on. If not, try another letter. This can be
optimized by deciding the letter to be replaced based on a table of
probabilities of letter consecutiveness and word length. For example,
the most common English word is "the". A combination that is 3 letters
and has "he" at the end is highly likely to be a "t", and there are a
limited number of other possibilities like "then", "they", "thee",
"she", "he", etc.
Of course, I'm only adding to the processor overhead, but it seems to me
that there is some breakeven point where the smaller dictionary size
makes up for the fact that the processor has to choose dictionaries and
do multiple guesses. Like I said, I'm sure you know more about this than
I do though.
It's been over 10 years on this stuff, so somebody tell me if I'm just
wrong.
|
Not Categorized, Please Help
|
|
