I need to work out a bunch of data with an algorithm i didn't choose. The algorithm has nothing to do with the "real" soundex algorithm (see Text::Soundex), even though it was called like that where i read it. So from now on i will baptize this algorithm Weirdex; this is described to receive a string in input and then:

• Strip chars outside [a-zA-Z] range
• Take all consonants except doubles (that should be reduced to one)
• Take just the first vowel

Examples:

Original	Weirdex
giulienk	gilnk
larry wall	larywl
etheroskedasticity	ethrskdstcty

My first coding of Weirdex is

sub weirdex {
    local $_ = shift;
    my ($weirdex, $vowel) = ('', 0);
    s/[^a-zA-Z]//g;
    for (split '') {
        if (/[aeiou]/i) {
            $vowel++ or $weirdex .= $_;
        } else {
            $weirdex .= $_ if substr($weirdex, -1, 1) ne $_;
        }
    }
    return $weirdex;
}
[download]

I'm interested in more elegant/efficient solutions, only regexp solutions and even Golf/Obfu ones.
Thank you.

$|=$_="1g2i1u1l2i4e2n0k",map{print"\7",chop;select$,,$,,$,,$_/7}m{..}g