This class knows how to read two treebank formats, the Penn format
and the Chomsky Normal Form (CNF) format. These formats differ in
how they handle terminal nodes. The Penn format places pre-terminal
part of speech tags in the left-hand position of a
parenthesis-delimited pair, just like it does non-terminal nodes.

The CNF format attaches pre-terminal tags to the word with an
underscore.

WWW: http://search.cpan.org/dist/Lingua-Treebank/