Since I'd guess the values for each field are substrings within the XML, your annotations could be directly on the xml document text with begin and end offsets addressing the substring. But if the substrings contain xml escape sequences such as & the annotation values would look a little ugly. I think another approach suggested was to extract the values from the xml, concatenate them together as a long string to form the document text, and re-apply the field structure by annotating each concatenated substring.
~Burn
