def convert_coq_to_nltk_type(coq_type):
"""
Given a coq_type specification such as:
Parameter _love : Entity -> Entity -> Prop.
return the equivalent NLTK type specification:
{'_love' : read_type('<e, <e, t>>')}
"""
assert isinstance(coq_type, str)
coq_type_list = coq_type.split()
assert len(coq_type_list) >= 4, 'Wrong coq_type format: %s' % coq_type
parameter, surface, colon = coq_type_list[:3]
assert parameter == 'Parameter' and colon == ':'
# This list contains something like ['Entity', '->', 'Prop', '->', 'Prop'...]
type_sig = coq_type_list[3:]
nltk_type_str = ' '.join(type_sig).rstrip('.').replace('->', ' ').replace(
'Entity', 'e').replace('Prop', 't').replace('Event', 'v')
if not nltk_type_str.startswith('(') or not nltk_type_str.endswith('('):
nltk_type_str = '(' + nltk_type_str + ')'
# Add pre-terminals (necessary for NLTK, if we convert to CNF).
nltk_type_str = re.sub(r'([evt])', r'(N \1)', nltk_type_str)
nltk_type_tree = tree_or_string(nltk_type_str)
nltk_type_tree.chomsky_normal_form(factor='right')
nltk_type_str = remove_labels_and_unaries(nltk_type_tree).replace(
'( ', '(').replace('(', '<').replace(')', '>').replace(' ', ',')
if len(type_sig) == 1:
nltk_type_str = nltk_type_str.strip('<>')
return {surface: read_type(nltk_type_str)}
def test_premise_case_var(self):
coq_line = 'H2 : _woman (Acc x1)'
expected_args = tree_or_string('(Acc x1)')
args = get_tree_pred_args(coq_line)
self.assertEqual(expected_args, args)
def parse_coq_line(coq_line):
try:
tree_args = tree_or_string('(' + coq_line + ')')
except ValueError:
tree_args = None
return tree_args
def test_premise_one_var(self):
coq_line = 'H2 : _kiss x1'
expected_args = tree_or_string('x1')
args = get_tree_pred_args(coq_line)
self.assertEqual(expected_args, args)