What dataset was the English tree parser trained on?

[Thilo Goetz]

This is still a todo in the documentation.  Does anybody know the
answer?  Was it trained on the Penn Treebank?  Thanks!

--Thilo