I'm posting this here because it's more (I hope) than a mere translation of the code. But we'll start with that:

sub min { # returns the smallest argument given
    my $min = shift;
    for (@_) {
        $min = $_ if $_ < $min;
    }
    return $min;
}

sub max { # returns the largest argument given
    my $max = shift;
    for (@_) {
        $max = $_ if $_ > $max;
    }
    return $max;
}

my %good; #number of occurences for each word in normal email
my %bad;  #number of occurences for each word in spam
my $ngood; # number of good messages
my $nbad;  # number of bad messages

sub score { # returns score for words given
    my @words = @_;#words in email
    my @probs; #probabilities for each word (not defined yet);

    for my $word (@words) {
        my $g = 2 * ($good{$word} || 0);
        my $b = $bad{$word} || 0;
        unless ($g + $b < 5) {
            push @probs,
                max( .01,
                     min( .99, 1 / min($b / $nbad) ),
                     min( 1, $g / $ngood ) + min( 1, $b / $nbad) );
        }
    }


    my $prod = 1;
    $prod *= $_  for (@probs);
    my $prod2 = 1;
    $prod2 *= $_ for (map {1 - $_ } @probs);
    return $prod / ($prod + $prod2);
}
[download]

After splitting up the message into different words, you call score with the words to retrieve the probability that the message is spam. Or at least I believe that's how it works. Oh, and by the way: the given code is a new dialect of lisp that's being developed named arc.

elusion : http://matt.diephouse.com