def load_peer_pyramids(
eval_tar: str, pyramids: Dict[str, Dict[str, Pyramid]]
) -> Dict[str, Dict[str, Dict[str, PyramidAnnotation]]]:
annotations = defaultdict(lambda: defaultdict(dict))
with tarfile.open(eval_tar, 'r') as tar:
for member in tar.getmembers():
if member.isfile() and member.name.startswith(
'UpdateSumm09_eval/manual/peers/'):
path = member.name.split('/')
filename = path[-1]
parts = filename.split('.')
instance_id = parts[0].split('-')[0].lower()
group = parts[0].split('-')[1]
summarizer_id = parts[-1]
# This directory only has peers
assert not summarizer_id.isalpha()
summarizer_type = 'peer'
pyramid = pyramids[instance_id][group]
xml = tar.extractfile(member).read().decode()
annotation = PyramidAnnotation.from_xml(
f'{instance_id}-{group}', summarizer_id, summarizer_type,
xml, pyramid)
if annotation:
annotations[instance_id][summarizer_id][group] = annotation
else:
print(
f'Annotation for {instance_id}-{group}, {summarizer_id} is `None`. Skipping'
)
return annotations
def load_peer_pyramids(
eval_tar: str,
pyramids: Dict[str,
Pyramid]) -> Dict[str, Dict[str, PyramidAnnotation]]:
annotations = defaultdict(dict)
with tarfile.open(eval_tar, 'r') as tar:
for member in tar.getmembers():
if member.isfile() and member.name.startswith('pans/'):
path = member.name.split('/')
filename = path[-1]
parts = filename.split('.')
instance_id = parts[1].lower()
summarizer_id = parts[5]
if summarizer_id.isalpha():
summarizer_type = 'reference'
else:
summarizer_type = 'peer'
pyramid = pyramids[instance_id]
xml = tar.extractfile(member).read().decode()
annotation = PyramidAnnotation.from_xml(
f'{instance_id}', summarizer_id, summarizer_type, xml,
pyramid)
if annotation:
annotations[instance_id][summarizer_id] = annotation
else:
print(
f'Annotation for {instance_id}, {summarizer_id} is `None`. Skipping'
)
return annotations
def load_main_annotations(
main_pyramid_tar: str,
pyramids: Dict[str,
Pyramid]) -> Dict[str, Dict[str, PyramidAnnotation]]:
annotations = defaultdict(dict)
with tarfile.open(main_pyramid_tar, 'r') as tar:
for member in tar.getmembers():
if member.isfile() and member.name.startswith(
'mainPyramidEval/allpans'):
path = member.name.split('/')
filename = path[-1]
parts = filename.split('.')
instance_id = parts[0].lower()
summarizer_id = parts[4]
# This directory only has peers
assert not summarizer_id.isalpha()
summarizer_type = 'peer'
pyramid = pyramids[instance_id]
xml = tar.extractfile(member).read().decode()
annotation = PyramidAnnotation.from_xml(
f'{instance_id}', summarizer_id, summarizer_type, xml,
pyramid)
if annotation:
annotations[instance_id][summarizer_id] = annotation
else:
print(
f'Annotation for {instance_id}, {summarizer_id} is `None`. Skipping'
)
return annotations
def load_peer_pyramids(
eval_tar: str, pyramids: Dict[str, Dict[str, Pyramid]]
) -> Dict[str, Dict[str, Dict[str, PyramidAnnotation]]]:
annotations = defaultdict(dict)
multiples = defaultdict(list)
with tarfile.open(eval_tar, 'r') as tar:
for member in tar.getmembers():
if member.isfile() and member.name.startswith('allpans/'):
path = member.name.split('/')
filename = path[-1]
parts = filename.split('.')
number = parts[0]
instance_id = parts[1].lower()
summarizer_id = parts[5]
# This directory only has peers
assert not summarizer_id.isalpha()
summarizer_type = 'peer'
pyramid = pyramids[instance_id]
xml = tar.extractfile(member).read().decode()
annotation = PyramidAnnotation.from_xml(
f'{instance_id}', summarizer_id, summarizer_type, xml,
pyramid)
if annotation:
if instance_id == 'd0631':
# This instance was annotated twice. We just take the first set
# for the pyramids files, but save both for a separate file
if number == '114':
annotations[instance_id][
summarizer_id] = annotation
multiples[summarizer_id].append(annotation)
else:
annotations[instance_id][summarizer_id] = annotation
else:
print(
f'Annotation for {instance_id}, {summarizer_id} is `None`. Skipping'
)
return annotations, multiples
def _get_scu_intersection(self, annotation: PyramidAnnotation, pyramid: Pyramid, index: int) -> Set[int]:
annotation_scus = annotation.get_scu_id_set()
reference_scus = pyramid.get_scu_id_set(index)
return annotation_scus & reference_scus