building a hierarchical classification out of flat and

i have problem with some "strange" type of data which i have to
convert to a hierarchical xml structure.

My source is a huge xml file which represents a decimal
classifikation. It contains so called documents, where each document
represents one node of the classification. Furthermore each documents
shows the direct parents of a node. It's a structure like this
(example taken from http://www.udcc.org):
...
<document>
	<tag1>3</tag1>
	<tag1a>Social Sciences</tag1a>
</document>
<document>
	<tag1>3</tag1>
	<tag1a>Social Sciences</tag1a>
	<tag2>32</tag2>
	<tag2a>Politics</tag2a>
</document>
<document>
	<tag1>3</tag1>
	<tag1a>Social Sciences</tag1a>
	<tag2>32</tag2>
	<tag2a>Politics</tag2a>
	<tag3>326</tag3>
	<tag3a>Slavery</tag3a>
</document>
...
As you can see there is no hierarchical information in it instead of
the names and the sequence of the tags. In my real data i have up to 9
levels, but not every time. My result should look like this (or
something similar):
...
<node id="3" name="Social Science">
  <node id="32" name="Politics">
     <node id="326" name="Slavery"/>
  </node>
</node>
...
I have simply no idea what to start with to archive this result. I
guess the first step would be to get rid of all those redundant
content, but i don't know how. And i even can't figure out how to
build the hierachichal structure the same time.