def load_special_combinators(filename):
for line in open(filename):
comment = line.find("#")
if comment > -1:
line = line[:comment]
line = line.strip()
if len(line) == 0: continue
items = line.split(" ")
head_is_left = items[0] == "l"
left = cat.parse(items[1])
right = cat.parse(items[2])
result = cat.parse(items[3])
res.append(SpecialCombinator(left, right, result, head_is_left))
return res
def load_unary(filename):
for line in open(filename):
comment = line.find("#")
if comment > -1:
line = line[:comment]
line = line.strip()
if len(line) == 0:
continue
items = line.split()
assert len(items) == 2
inp = cat.parse(items[0])
out = cat.parse(items[1])
if res.has_key(inp):
res[inp].append(out)
else:
res[inp] = [out]
return res
def parse_leaf(self):
self.word_id += 1
self.check("{")
cate = self.next(" ")[1:].encode("utf-8")
cate = cate[:cate.find("_")]
cate = cat.parse(cate)
word = self.next("}")[:-1].split("/")[0]
return Leaf(word, cate, self.word_id)
def parse_tree(self):
self.check("{")
op = self.next(" ")
op = combinators[op[1:]]
cate = cat.parse(self.next(" ").encode("utf-8"))
self.check("{")
children = []
while self.peek() != "}":
children.append(self.next_node())
if self.peek() == " ":
self.next(" ")
self.next("}")
left_is_head = True if len(children) == 1 else \
op.head_is_left(children[0].cat, children[1].cat)
return Tree(cate, left_is_head, children, op)
def can_apply(self, left, right):
# Comments from easyCCG:
# * Don't start making weird ,\, categories...
# * Improves coverage of C&C evaluation script.
# Categories can just conjoin first, then type-raise.
# (not self.right.is_type_raised)
# * Blocks noun conjunctions, which should normally be NP conjunctions.
# In a better world, conjunctions would have categories like (NP\NP/NP.
# Doesn't affect F-scopes, but makes output semantically nicer.
#
# issues related to C&C evaluation script?
# """C&C evaluation script does't let you do this, for some reason"""
if cat.parse("NP\\NP").matches(right):
return False
return (left == cat.CONJ or \
left == cat.COMMA or \
left == cat.SEMICOLON) and \
not right.is_punct and \
not right.is_type_raised and \
not (not right.is_functor and right.type == "N")