It is probably best to store both the questions and answers in the same doc, and then break out the questions and answers using different filters. Here is an example approach:

<?xml version="1.0"?>
<data>
  <question id="1">
    <qtext>What does KB stand for?</qtext>
    <choices>
      <choice>Kilo Bits</choice>
      <choice>Key Board</choice>
      <choice>Kilo Bytes</choice>
      <choice>None</choice>
    </choices>
    <answer>2</answer>
    <answer>3</answer>
  </question>
  <question id="2">
    <qtext>CPU stands for</qtext>
    <choices>
      <choice>Central Processing Unit</choice>
      <choice>Central Power Unit</choice>
      <choice>Core Processing Unit</choice>
      <choice>Core Power Unit</choice>
    </choices>
    <answer>1</answer>
    <answer>4</answer>
  </question>
  <question id="3">
    <qtext>1 KB equals</qtext>
    <choices>
      <choice>1000 Bytes</choice>
      <choice>1024 Bytes</choice>
      <choice>1000 Bits</choice>
      <choice>1024 Bits</choice>
      <choice>Nothing</choice>
    </choices>
    <answer>2</answer>
    <answer>1</answer>
  </question>
</data>
[download]

Then use any XML parser, XML::Parser, XML::Twig, etc. to extract the <choices> and <answer> parts separately. Yould could also probably do this using XSLT, too, if you wanted to reamin XML pure. But I would go for the perl-based solutions first.