About:

This is text data from the Wall Street Journal. It was parsed for sentence
breaks, punctuation, abbreviations, capitalization, etc. 

In wsj.xml, each element S represents a sentence, and words/punctuation of
the sentence are listed in order as subelements T.

The file wsj2.xml is a reformatting of the same data, where each element S
represents a sentence, and consecutive words are denoted by the elements T
and W. For example, the sentence "Roses are red." would be represented by
the element

  <S>
    <T token="START"><W token="roses"/></T>
    <T token="roses"><W token="are"/></T>
    <T token="are"><W token="red"/></T>
    <T token="red"><W token="PERIOD"/></T>
    <T token="PERIOD"><W token="EOS"/></T>
  </S>

Please don't ask why it's all weird like that.