Nice++
Did a benchmark (I'll post later with some tweaks) and it seems that for a few terms, the match/span is about twice as fast then the double regex, with more terms it goes down to about 70% faster.

Simply splitting without actually doing any processing

    my ($text,@words) = @_;
    my @text = split /\s+/, $text;
    my @results;
    for my $t (@text){
      for my $w (@words){
        push @results if $t eq $w;
      }
    }
[download]

is about 46% slower then the pattern match/span solution.