I always think that parsing HTML with regexp's is bad and go for something like HTML::Parser or HTML::TreeBuilder everytime. This may or may not be an overkill for your situation.

use HTML::Parser;

my $keyword = quotemeta 'match';

sub highlight {
    my $text = shift;
    $text =~ s|\b($keyword)\b|<span style="background: #FFFFFF">$1</sp
+an>|g;
    return $text;
}

my $html = q{
    <p id="match"><b>this is html match this</b><u value="don't match 
+this">text html blah</u></p>
};

my $p = new HTML::Parser(api_version => 3, 
    handlers => {
        text=> [ sub { print highlight(shift) }, 'text'],
    default => [sub { print shift }, 'text'],
  });

$p->parse($html);
[download]

gav^