The text node is notified to the characters() event in your ContentHandler, which can modify the content any way it likes.
The parser is allowed to make multiple calls on characters(), dividing the text into fragments any way it likes, so you should buffer it up and process it the next time endElement or startElement is called.
Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference