Mark T.B. Carroll <Mark.Carroll@aetion.com> wrote: ssecorp <circularfunc@gmail.com> writes: classify good bad test = (countall bad test) / ((countall bad test) + (countall good test)) here I do (countall bad test) twice. will this somehow be done only once though because the compiler is smart enough to figure that out? With GHC often it won't. I know it does some

Below is a simple spamfilter I have written in Haskell. It is very slow but effective and trainable. I will make a better version tomorrow, right now I do a lot of going over lists many times and probably have some exponential functions in there. I will also choose better names for the functions and add type declarations. However it works and does a good job after just a few emails. It is

"Mark T.B. Carroll" <Mark.Carroll@Aetion.com> writes: counter (s:set) (list) tuple | (set) /= [] = counter set (list) (tuple ++ [(s, occur s (list))]) | (set) == [] = (tuple ++ [(s, occur s (list))]) countall (s:set) (list) = counter (s:set) (list) [] You mostly don't need the parentheses where you don't have a `:'. Also, `null' is a more typical`==[]' test

aha! occ x (a:ab) count | x == a && ab /= [] = occ x (ab) (count + 1) | x /= a && ab /= [] = occ x (ab) count | x == a && ab == [] = count + 1 | otherwise = count occur x (a:ab) = occ x (a:ab) 0 counter (s:set) (list) tuple | (set) /= [] = counter set (list) (tuple ++ [(s, occur s (list))]) | (set) == [] = (tuple ++ [(s, occur s (list