def get(self, request, item_id, item_type="docket"):
if item_type == "document":
doc = Doc.objects.only("docket_id").get(id=item_id)
self.kwargs['docket_id'] = doc.docket_id
else:
self.kwargs['docket_id'] = item_id
hierarchy = self.corpus.hierarchy()
out = {
'docket_teaser': {
'0.5': {'count': self._count_clusters(hierarchy, 0.5)},
'0.8': {'count': self._count_clusters(hierarchy, 0.8)}
}
}
if item_type == 'document':
out['document_teaser'] = None
docs = self.corpus.docs_by_metadata('document_id', item_id)
if docs:
out['document_teaser'] = {}
doc_id = docs[0]
cluster05 = find_doc_in_hierarchy(hierarchy, doc_id, 0.5)
if cluster05:
out['document_teaser'] = {'0.5': {'count': cluster05['size'], 'id': doc_id}}
cluster08 = find_doc_in_hierarchy(hierarchy, doc_id, 0.8)
if cluster08:
out['document_teaser']['0.8'] = {'count': cluster08['size'], 'id': doc_id}
return Response(out)
def get(self, request, docket_id):
docket = Docket.objects.get(id=docket_id)
hierarchy = self.corpus.hierarchy(
request.GET.get('require_summaries', "").lower() == "true")
total_clustered = sum([cluster['size'] for cluster in hierarchy])
out = {
'cluster_hierarchy':
sorted(hierarchy, key=lambda x: x['size'], reverse=True),
'stats': {
'clustered':
total_clustered,
'unclustered':
docket.stats['count'] -
total_clustered if 'count' in docket.stats else None,
'date_range':
docket.stats['date_range']
if 'date_range' in docket.stats else None
},
'prepopulate':
None
}
# populate agency info
agency = docket.agency or docket.id.split("-")[0]
if agency:
agency_meta = list(Agency.objects(id=agency).only("name"))
if agency_meta:
out['stats']['agency'] = {
'id': agency,
'name': agency_meta[0].name,
'url': '/agency/%s' % agency
}
else:
out['stats']['agency'] = None
# choose a cluster and document to prepopulate if one hasn't been requested
prepop = int(request.GET.get('prepopulate_document', -1))
if prepop > -1:
pp_cluster = find_doc_in_hierarchy(hierarchy, prepop, self.cutoff)
if pp_cluster:
out['prepopulate'] = {
'document': prepop,
'cluster': pp_cluster['name'],
'cutoff': self.cutoff
}
if not out['prepopulate'] and out['stats']['clustered'] > 0:
pp_cluster = find_doc_in_hierarchy(
hierarchy, out['cluster_hierarchy'][0]['name'],
out['cluster_hierarchy'][0]['cutoff'])
out['prepopulate'] = {
'document': pp_cluster['members'][0],
'cluster': pp_cluster['name'],
'cutoff': out['cluster_hierarchy'][0]['cutoff']
}
remove_members(out['cluster_hierarchy'])
return Response(out)
def get(self, request, docket_id):
docket = Docket.objects.get(id=docket_id)
hierarchy = self.corpus.hierarchy(request.GET.get('require_summaries', "").lower()=="true")
total_clustered = sum([cluster['size'] for cluster in hierarchy])
out = {
'cluster_hierarchy': sorted(hierarchy, key=lambda x: x['size'], reverse=True),
'stats': {
'clustered': total_clustered,
'unclustered': docket.stats['count'] - total_clustered if 'count' in docket.stats else None,
'date_range': docket.stats['date_range'] if 'date_range' in docket.stats else None
},
'prepopulate': None
}
# populate agency info
agency = docket.agency
if agency:
agency_meta = list(Agency.objects(id=agency).only("name"))
if agency_meta:
out['stats']['agency'] = {
'id': agency,
'name': agency_meta[0].name,
'url': '/agency/%s' % agency
}
else:
out['stats']['agency'] = None
# choose a cluster and document to prepopulate if one hasn't been requested
prepop = int(request.GET.get('prepopulate_document', -1))
if prepop > -1:
pp_cluster = find_doc_in_hierarchy(hierarchy, prepop, self.cutoff)
if pp_cluster:
out['prepopulate'] = {
'document': prepop,
'cluster': pp_cluster['name'],
'cutoff': self.cutoff
}
if not out['prepopulate'] and out['stats']['clustered'] > 0:
pp_cluster = find_doc_in_hierarchy(hierarchy, out['cluster_hierarchy'][0]['name'], out['cluster_hierarchy'][0]['cutoff'])
out['prepopulate'] = {
'document': pp_cluster['members'][0],
'cluster': pp_cluster['name'],
'cutoff': out['cluster_hierarchy'][0]['cutoff']
}
remove_members(out['cluster_hierarchy'])
return Response(out)
def get(self, request, docket_id, cluster_id):
cluster_id = int(cluster_id)
h = self.corpus.hierarchy()
cluster = find_doc_in_hierarchy(h, cluster_id, self.cutoff)
# consider caching for very large clusters
_metadatas = lambda: dict(self.corpus.doc_metadatas(cluster['members']))
members = tuple(cluster['members'])
if len(members) > 1000:
key = 'sparerib_api.clustering.cluster-%s-%s-%s' % (docket_id, cluster_id, hash(members))
metadatas = cache.get(key)
if not metadatas:
metadatas = _metadatas()
cache.set(key, metadatas)
else:
# it's little; don't cache
metadatas = _metadatas()
return Response({
'id': cluster['name'],
'documents': [{
'id': doc_id,
'title': metadatas[doc_id]['title'],
'submitter': ', '.join([metadatas[doc_id][field] for field in ['submitter_name', 'submitter_organization'] if field in metadatas[doc_id] and metadatas[doc_id][field]])
} for doc_id in cluster['members']]
})
def get(self, request, item_id, item_type="docket"):
if item_type == "document":
doc = Doc.objects.only("docket_id").get(id=item_id)
self.kwargs['docket_id'] = doc.docket_id
else:
self.kwargs['docket_id'] = item_id
hierarchy = self.corpus.hierarchy()
out = {
'docket_teaser': {
'0.5': {
'count': self._count_clusters(hierarchy, 0.5)
},
'0.8': {
'count': self._count_clusters(hierarchy, 0.8)
}
}
}
if item_type == 'document':
out['document_teaser'] = None
docs = self.corpus.docs_by_metadata('document_id', item_id)
if docs:
out['document_teaser'] = {}
doc_id = docs[0]
cluster05 = find_doc_in_hierarchy(hierarchy, doc_id, 0.5)
if cluster05:
out['document_teaser'] = {
'0.5': {
'count': cluster05['size'],
'id': doc_id
}
}
cluster08 = find_doc_in_hierarchy(hierarchy, doc_id, 0.8)
if cluster08:
out['document_teaser']['0.8'] = {
'count': cluster08['size'],
'id': doc_id
}
return Response(out)
def get(self, request, docket_id, cluster_id, document_id):
document_id = int(document_id)
cluster_id = int(cluster_id)
h = self.corpus.hierarchy()
cluster = find_doc_in_hierarchy(h, cluster_id, self.cutoff)['members']
doc = self.corpus.doc(document_id)
text = doc['text']
raw_phrases = self.corpus.phrase_overlap(document_id, cluster)
frequencies = numpy.zeros(len(text), 'l')
for phrase in raw_phrases.values():
for occurrence in phrase['indexes']:
frequencies[occurrence.start:occurrence.end] = numpy.maximum(
frequencies[occurrence.start:occurrence.end],
phrase['count'])
freq_ranges = [(f[0], len(list(f[1])))
for f in itertools.groupby(frequencies)]
cluster_size = float(len(cluster))
components = []
cursor = 0
for fr in freq_ranges:
components.append((fr[0], text[cursor:cursor + fr[1]]))
cursor += fr[1]
html = ''.join([
'<span style="background-color:rgba(160,211,216,%s)">%s</span>' %
(round(p[0] / cluster_size, 2), p[1]) for p in components
])
html = html.replace("\n", "<br />")
return Response({
'metadata': {
'title':
doc['metadata'].get('title', None),
'submitter':
', '.join([
doc['metadata'][field]
for field in ['submitter_name', 'submitter_organization']
if field in doc['metadata'] and doc['metadata'][field]
]),
'document_id':
doc['metadata'].get('document_id', None),
},
'frequency_html': html,
'truncated': len(doc['text']) == 10000
})
def get(self, request, docket_id, cluster_id):
cluster_id = int(cluster_id)
h = self.corpus.hierarchy()
cluster = find_doc_in_hierarchy(h, cluster_id, self.cutoff)
metadatas = dict(self.corpus.doc_metadatas(cluster['members']))
return Response({
'id': cluster['name'],
'documents': [{
'id': doc_id,
'title': metadatas[doc_id]['title'],
'submitter': ', '.join([metadatas[doc_id][field] for field in ['submitter_name', 'submitter_organization'] if field in metadatas[doc_id] and metadatas[doc_id][field]])
} for doc_id in cluster['members']]
})
def get(self, request, docket_id, cluster_id):
cluster_id = int(cluster_id)
h = self.corpus.hierarchy()
cluster = find_doc_in_hierarchy(h, cluster_id, self.cutoff)
# consider caching for very large clusters
_metadatas = lambda: dict(self.corpus.doc_metadatas(cluster['members'])
)
members = tuple(cluster['members'])
if len(members) > 1000:
key = 'sparerib_api.clustering.cluster-%s-%s-%s' % (
docket_id, cluster_id, hash(members))
metadatas = cache.get(key)
if not metadatas:
metadatas = _metadatas()
cache.set(key, metadatas)
else:
# it's little; don't cache
metadatas = _metadatas()
return Response({
'id':
cluster['name'],
'documents': [{
'id':
doc_id,
'title':
metadatas[doc_id]['title'],
'submitter':
', '.join([
metadatas[doc_id][field]
for field in ['submitter_name', 'submitter_organization']
if field in metadatas[doc_id] and metadatas[doc_id][field]
])
} for doc_id in cluster['members']]
})
def get(self, request, docket_id, cluster_id, document_id):
document_id = int(document_id)
cluster_id = int(cluster_id)
h = self.corpus.hierarchy()
cluster = find_doc_in_hierarchy(h, cluster_id, self.cutoff)['members']
doc = self.corpus.doc(document_id)
text = doc['text']
raw_phrases = self.corpus.phrase_overlap(document_id, cluster)
frequencies = numpy.zeros(len(text), 'l')
for phrase in raw_phrases.values():
for occurrence in phrase['indexes']:
frequencies[occurrence.start:occurrence.end] = numpy.maximum(frequencies[occurrence.start:occurrence.end], phrase['count'])
freq_ranges = [(f[0], len(list(f[1]))) for f in itertools.groupby(frequencies)]
cluster_size = float(len(cluster))
components = []
cursor = 0
for fr in freq_ranges:
components.append((fr[0], text[cursor:cursor + fr[1]]))
cursor += fr[1]
html = ''.join(['<span style="background-color:rgba(160,211,216,%s)">%s</span>' % (round(p[0]/cluster_size, 2), p[1]) for p in components])
html = html.replace("\n", "<br />")
return Response({
'metadata': {
'title': doc['metadata'].get('title', None),
'submitter': ', '.join([doc['metadata'][field] for field in ['submitter_name', 'submitter_organization'] if field in doc['metadata'] and doc['metadata'][field]]),
'document_id': doc['metadata'].get('document_id', None),
},
'frequency_html': html,
'truncated': len(doc['text']) == 10000
})
