tree_parser.py

"""
Parser for Penn WSJ trees.  

It is basically a dumbed-down version of the S-expression parser provided as
an example of the pyparsing module.  This example is available at:
    http://pyparsing.wikispaces.com/file/view/sexpParser.py
"""
import pprint
import pyparsing

# Define punctuation literals
LPAR, RPAR, LBRK, RBRK, LBRC, RBRC = map(pyparsing.Suppress, "()[]{}")

# The symbols a token can contain
token = pyparsing.Word(pyparsing.alphanums + "-./_:;*+=!<>@&`',?%#$\\")

display = LBRK + token + RBRK
string_ = pyparsing.Optional(display) + token

sexp = pyparsing.Forward()
sexpList = pyparsing.Group(LPAR + pyparsing.ZeroOrMore(sexp) + RPAR)
sexp << (string_ | sexpList)

def get_tree(text_tree):
    """Returns a nested list of strings representing a textual parse tree.

    The tokens are not interpreted.  For example, numerical-valued leaf nodes
    are still represented as strings in the tree.
    
    Args:
        text_tree: a string representation of a single parse tree. These are
            defined by the Penn WSJ database.  A sample input is:
                "(TOP (INTJ (UH damn) (. !)) )"
    
    Returns:
        The parse tree of the input, represented in a list datastructure.  For
        example:
            ['TOP', ['INTJ', ['UH, 'damn'], ['.', '!']]]
    """
    return sexp.parseString(text_tree, parseAll=True).asList()[0]