Tidy HTML inserted by copy/pasting from Microsoft office (PHP)

function.tidy_office_html.php

// Regexps courtesy of 1st class media
// http://www.1stclassmedia.co.uk/developers/clean-ms-word-formatting.php
function tidy_office_html($str) {
	$replacements = array(
		'/<!--.*?-->/s' => '',
		'/<o:p>\s*<\/o:p>/s' => '',
		'/<o:p>.*?<\/o:p>/s' => "&nbsp;",
		'/\s*mso-[^:]+:[^;"]+;?/i' => '',
		'/\s*MARGIN: 0cm 0cm 0pt\s*;/i' => '',
		'/\s*MARGIN: 0cm 0cm 0pt\s*"/i' => '',
		'/\s*TEXT-INDENT: 0cm\s*;/i' => '',
		'/\s*TEXT-INDENT: 0cm\s*"/i' => '',
		'/\s*TEXT-ALIGN: [^\s;]+;?"/i' => '',
		'/\s*PAGE-BREAK-BEFORE: [^\s;]+;?"/i' => '',
		'/\s*FONT-VARIANT: [^\s;]+;?"/i' => '',
		'/\s*tab-stops:[^;"]*;?/i' => '',
		'/\s*tab-stops:[^"]*/i' => '',
		'/\s*face="[^"]*"/i' => '',
		'/\s*face=[^ >]*/i' => '',
		'/\s*FONT-FAMILY:[^;"]*;?/i' => '',
		'/<(\w[^>]*) class=([^ |>]*)([^>]*)/i' => "<$1$3",
		'/<(\w[^>]*) style="([^\"]*)"([^>]*)/i' => "<$1$3",
		'/\s*style="\s*"/i' => '',
		'/<SPAN\s*[^>]*>\s*&nbsp;\s*<\/SPAN>/i' => '&nbsp;',
		'/<SPAN\s*[^>]*><\/SPAN>/i' => '',
		'/<(\w[^>]*) lang=([^ |>]*)([^>]*)/i' => "<$1$3",
		'/<SPAN\s*>(.*?)<\/SPAN>/i' => '$1',
		'/<FONT\s*>(.*?)<\/FONT>/i' => '$1',
		':<p>&nbsp;</p>:i' => '',
		'/<\\?\?xml[^>]*>/i' => '',
		'/<\/?\w+:[^>]*>/i' => '',
		'/<([^\s>]+)[^>]*>\s*<\/\1>/s' => '',
	);
	foreach ($replacements as $pattern => $replacement) {
		$str = preg_replace($pattern, $replacement, $str);
	}
	return $str;
}

The reason behind the foreach loop is to prevent a duplicate internal loop within array_keys() and array_values() as in

return preg_replace(array_keys($replacements), array_values($replacements), $str);

I did not benchmark this though. For even more optimization (but decreased readability), the two arrays should be defined separately to get rid of the need for splitting them into keys and values.