def _element_2_syntactic_node(cls, element):
cat = element.attrib["cat"]
if element.tag == "lf":
terminal_node = SyntacticNode(cat,
pos=element.attrib["pos"],
lemma=element.attrib["lemma"],
word=element.attrib["word"])
return terminal_node
else:
node = SyntacticNode(cat)
for child_element in element:
child_node = cls._element_2_syntactic_node(child_element)
node.add_child(child_node)
return node
def read_tree(cls, string):
"""Read a SyntacticTree from a string
Args:
string: the input string
For a PSG tree, the format of the tree is similar to a Penn Treebank
tree, without newline characters:
e.g. (S (NP (NNP Mary)) (VP (VBZ loves) (NP (NNP John))))
Returns:
the SyntacticTree represented by the input string
"""
# first, remove the brackets ()
string = string.strip()
if len(string) == 0:
raise ValueError("empty string cannot be a Synstactic Tree")
if string[0] == "(" and string[-1] == ")":
string = string[1:-1]
# split the string with blank character
# if the string has exactly or fewer than one element, it cannot be a
# tree
# if it has two elements, it must be a tree with a terminal node as the
# root
# if it has more than two elements, take the first element as the root
# and other elements as the branches
elements = string.split()
if len(elements) <= 1 or elements[1] == "":
raise ValueError("%s cannot be a tree or subtree" %string)
else:
if len(elements) == 2:
# TODO: if the label comes from CCG parser, turn [] into ()
root = cls._read_terminal_node(elements)
else:
branch_string = " ".join(elements[1:])
root = SyntacticNode(elements[0])
branches = cls._read_branches(branch_string)
for branch in branches:
root.add_child(branch._root)
return SyntacticTree(root)
