mirror of https://github.com/doccano/doccano.git
pythondatasetsactive-learningtext-annotationdatasetnatural-language-processingdata-labelingmachine-learningannotation-tool
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
107 lines
3.7 KiB
107 lines
3.7 KiB
from django.db.models import Count, Manager
|
|
|
|
|
|
class AnnotationManager(Manager):
|
|
|
|
def calc_label_distribution(self, examples, members, labels):
|
|
"""Calculate label distribution.
|
|
|
|
Args:
|
|
examples: example queryset.
|
|
members: user queryset.
|
|
labels: label queryset.
|
|
|
|
Returns:
|
|
label distribution per user.
|
|
|
|
Examples:
|
|
>>> self.calc_label_distribution(examples, members, labels)
|
|
{'admin': {'positive': 10, 'negative': 5}}
|
|
"""
|
|
distribution = {member.username: {label.text: 0 for label in labels} for member in members}
|
|
items = self.filter(example_id__in=examples)\
|
|
.values('user__username', 'label__text')\
|
|
.annotate(count=Count('label__text'))
|
|
for item in items:
|
|
username = item['user__username']
|
|
label = item['label__text']
|
|
count = item['count']
|
|
distribution[username][label] = count
|
|
return distribution
|
|
|
|
def get_labels(self, label, project):
|
|
if project.collaborative_annotation:
|
|
return self.filter(example=label.example)
|
|
else:
|
|
return self.filter(example=label.example, user=label.user)
|
|
|
|
def can_annotate(self, label, project) -> bool:
|
|
raise NotImplementedError('Please implement this method in the subclass')
|
|
|
|
def filter_annotatable_labels(self, labels, project):
|
|
return [label for label in labels if self.can_annotate(label, project)]
|
|
|
|
|
|
class CategoryManager(AnnotationManager):
|
|
|
|
def can_annotate(self, label, project) -> bool:
|
|
is_exclusive = project.single_class_classification
|
|
categories = self.get_labels(label, project)
|
|
if is_exclusive:
|
|
return not categories.exists()
|
|
else:
|
|
return not categories.filter(label=label.label).exists()
|
|
|
|
|
|
class SpanManager(AnnotationManager):
|
|
|
|
def can_annotate(self, label, project) -> bool:
|
|
overlapping = getattr(project, 'allow_overlapping', False)
|
|
spans = self.get_labels(label, project)
|
|
if overlapping:
|
|
return True
|
|
for span in spans:
|
|
if span.is_overlapping(label):
|
|
return False
|
|
return True
|
|
|
|
|
|
class ExampleManager(Manager):
|
|
|
|
def bulk_create(self, objs, batch_size=None, ignore_conflicts=False):
|
|
super().bulk_create(objs, batch_size=batch_size, ignore_conflicts=ignore_conflicts)
|
|
uuids = [data.uuid for data in objs]
|
|
examples = self.in_bulk(uuids, field_name='uuid')
|
|
return [examples[uid] for uid in uuids]
|
|
|
|
|
|
class ExampleStateManager(Manager):
|
|
|
|
def count_done(self, examples, user=None):
|
|
if user:
|
|
queryset = self.filter(example_id__in=examples, confirmed_by=user)
|
|
else:
|
|
queryset = self.filter(example_id__in=examples)
|
|
return queryset.distinct().values('example').count()
|
|
|
|
def measure_member_progress(self, examples, members):
|
|
done_count = self.filter(example_id__in=examples)\
|
|
.values('confirmed_by__username')\
|
|
.annotate(total=Count('confirmed_by'))
|
|
response = {
|
|
'total': examples.count(),
|
|
'progress': [
|
|
{
|
|
'user': obj['confirmed_by__username'],
|
|
'done': obj['total']
|
|
} for obj in done_count
|
|
]
|
|
}
|
|
members_with_progress = {o['confirmed_by__username'] for o in done_count}
|
|
for member in members:
|
|
if member.username not in members_with_progress:
|
|
response['progress'].append({
|
|
'user': member.username,
|
|
'done': 0
|
|
})
|
|
return response
|