Ok, I have written a program that you can customize. Just set $targetto what you want to wipe out of text segments... for some reason the   is not matching when I switch $target to it and I dont know why.

But as you can see, the "hi" was removed... I will ask on the libwww-perl list about this.

use strict;
use warnings;
use HTML::TreeBuilder;

my $html = '<h1>Blah</h1>&nbsp;hithere&nbsp;<p>&nbsp;<br>
<h2>Blah</h2>';

my $element_root = HTML::TreeBuilder->new_from_content($html);

$element_root->objectify_text;

$element_root->dump;
print "\n";

my $target;
$target = '&nbsp;';
$target = 'hi';

my @text = $element_root->look_down('_tag' => '~text');
for my $text_node (@text) {
  my $tmp = $text_node->attr('text');
  warn $tmp;
  if ($tmp =~ m!$target!) {
    warn 'here';
    $tmp =~ s!$target!!g;
    $text_node->attr(text => $tmp);
  }
}

print "\n";
$element_root->dump;
[download]