I'd build a hash of all the words in the file and then remove the common words. Use each word as the key to the hash, it's value doesn't matter.

open my $IN, "<", "myfile.txt" or die $!;

my %seen;
while ( <$IN> ) {
    $seen{$_}++ for split;
}

delete $seen{$_} for qw/all my common words/;

my @metadata = keys %seen;
[download]