Split element with mixed content

Hello

What is the best method for splitting an element that has mixed content?

INPUT
-----
<root>
    <p>The quick <b>brown</b> fox jumped over the lazy dog.</p>
</root>

In the above example I want to split on the word 'over'.

OUTPUT
-----
<root>
    <p>The quick <b>brown</b> fox jumped </p>
    <p>over the lazy dog.</p>
</root>

Using analyze-string and tokenize only gives me the string value of the
<b> element. Am I missing something very obvious?

Thanks.

--
Kevin