def make_annotation(self, pronoun, a_coref, b_coref):
annotation = Annotation()
annotation.name_a_coref = a_coref
annotation.name_b_coref = b_coref
gender = PRONOUNS.get(pronoun.lower(), Gender.UNKNOWN)
assert gender != Gender.UNKNOWN
annotation.gender = gender
return annotation
def read_annotations(filename, is_gold):
"""Reads coreference annotations for the examples in the given file.
Args:
filename: Path to .tsv file to read.
is_gold: Whether or not we are reading the gold annotations.
Returns:
A dict mapping example ID strings to their Annotation representation. If
reading gold, 'Pronoun' field is used to determine gender.
"""
def is_true(value):
if value.lower() == 'true':
return True
elif value.lower() == 'false':
return False
else:
print('Unexpected label!', value,"in file:",filename)
return None
fieldnames = GOLD_FIELDNAMES if is_gold else SYSTEM_FIELDNAMES
annotations = defaultdict(Annotation)
with open(filename, 'rU') as f:
reader = csv.DictReader(f, fieldnames=fieldnames, delimiter='\t')
# Skip the header line in the gold data
if is_gold:
next(reader, None)
first_row = True
for row in reader:
if first_row:
first_row = False
continue
example_id = row['ID']
if example_id in annotations:
print('Multiple annotations for', example_id)
continue
annotations[example_id].name_a_coref = is_true(row['A-coref'])
annotations[example_id].name_b_coref = is_true(row['B-coref'])
if is_gold:
gender = PRONOUNS.get(row['Pronoun'].lower(), Gender.UNKNOWN)
assert gender != Gender.UNKNOWN, row
annotations[example_id].gender = gender
return annotations
