def process_document(self, document: Document, params: Dict[str, Any]):
terms_index_name = params.get('terms_index', 'umls_terms')
terms = document.labels[terms_index_name]
negation_triggers = document.labels['negation_triggers']
all_deps = []
all_upos_tags = []
sentences = []
sentence_texts = []
for sentence in document.labels['sentences']:
if len(terms.inside(sentence)) == 0 or len(
negation_triggers.inside(sentence)) == 0:
continue
sentences.append(sentence)
sentence_texts.append(sentence.text)
stanza_doc = self.nlp(sentence_texts)
for (sentence, stanza_sentence) in zip(sentences,
stanza_doc.sentences):
sentence_deps, sentence_upos_tags = stanza_deps_and_upos_tags(
sentence, stanza_sentence)
all_deps.extend(sentence_deps)
all_upos_tags.extend(sentence_upos_tags)
document.add_labels('dependencies', all_deps)
document.add_labels('upos_tags', all_upos_tags)
def process_document(self, document: mtap.Document, params: Dict[str,
Any]):
referenced = [
mtap.GenericLabel(0, 1),
mtap.GenericLabel(1, 2),
mtap.GenericLabel(2, 3),
mtap.GenericLabel(3, 4)
]
# references can be a map of strings to labels
with document.get_labeler('map_references') as label_map_references:
label_map_references(0,
4,
ref={
'a': referenced[0],
'b': referenced[1],
'c': referenced[2],
'd': referenced[3]
})
# references can be a list of labels
with document.get_labeler('list_references') as label_list_references:
label_list_references(0, 2, ref=[referenced[0], referenced[1]])
label_list_references(2, 3, ref=[referenced[2], referenced[3]])
# references can be direct
with document.get_labeler('references') as label_references:
label_references(0, 2, a=referenced[0], b=referenced[1])
label_references(2, 3, a=referenced[2], b=referenced[3])
# referenced labels don't need to be added via "addLabels" or "Labeler.close" before label
# indices that reference them.
# The Document will delay uploading any label indices to the server until they are.
document.add_labels('referenced', referenced)
def copy_document(event: Event,
source_document_name: str,
target_document_name: str,
index_names: Sequence[str] = ...):
"""Copies one document to another on the same event.
Parameters
----------
event: Event
The event.
source_document_name: str
The source document name.
target_document_name: str
The target document name.
index_names: Sequence[str]
If specified will only copy the specified label indices, by default all indices will be
copied.
"""
source_document = event.documents[source_document_name]
target_document = Document(target_document_name, text=source_document.text)
event.add_document(target_document)
if index_names is ...:
info = source_document.get_label_indices_info()
index_names = [i.index_name for i in info]
for index_name in index_names:
index = source_document.get_label_index(index_name)
target_document.add_labels(index_name, index, distinct=index.distinct)
def process_document(self, document: Document, params: Dict[str, Any]):
label_trigger = document.get_labeler('negation_triggers')
with label_trigger:
for sentence in document.get_label_index('sentences'):
triggers = self.negex.detect_negex_triggers(sentence.text)
for start_index, end_index, tags in triggers:
label_trigger(sentence.start_index + start_index,
sentence.start_index + end_index,
tags=tags)
def process_document(self,
document: Document,
params: Dict[str, Any]):
sentences = document.labels['sentences']
sentences_text = []
for sentence in sentences:
sentences_text.append(sentence.text)
stanza_doc = self.nlp(sentences_text)
all_deps = []
all_upos_tags = []
for stanza_sentence, sentence in zip(stanza_doc.sentences, sentences):
dependencies = {}
stanza_dependencies = stanza_sentence.dependencies
stanza_dependencies = list(stanza_dependencies)
i = 0
while len(stanza_dependencies) > 0:
i += 1
if i > MAX_ITER:
raise ValueError(
'Maximum Iterations reached while processing dependency graph.')
head, deprel, dep = stanza_dependencies.pop()
head_id = int(head.id)
if head_id == 0:
head_dep_label = None
else:
try:
head_dep_label = dependencies[head_id]
except KeyError:
stanza_dependencies.insert(0, (head, deprel, dep))
continue
token_begin = sentence.start_index + dep.parent.start_char - stanza_sentence.tokens[
0].start_char
token_end = sentence.start_index + dep.parent.end_char - stanza_sentence.tokens[
0].start_char
dep_label = GenericLabel(token_begin, token_end, head=head_dep_label, deprel=deprel)
dep_label.reference_cache['dependents'] = []
dependencies[int(dep.id)] = dep_label
if head_dep_label is not None:
head_dep_label.dependents.append(dep_label)
all_deps.append(dep_label)
for word in stanza_sentence.words:
token = word.parent
token_begin = sentence.start_index + token.start_char - stanza_sentence.tokens[
0].start_char
token_end = sentence.start_index + token.end_char - stanza_sentence.tokens[
0].start_char
all_upos_tags.append(GenericLabel(token_begin, token_end, tag=word.upos))
document.add_labels('dependencies', all_deps)
document.add_labels('upos_tags', all_upos_tags)
def test_labeler_distinct_and_type_id_raises(mocker):
with pytest.raises(ValueError):
client = mocker.Mock(EventsClient)
event = Event(event_id='1', client=client)
document = Document(
document_name='plaintext',
text='The quick brown fox jumped over the lazy dog.',
event=event)
document.get_labeler('index',
distinct=True,
label_adapter=DistinctGenericLabelAdapter)
def test_yml_serializer():
event = Event(event_id='1')
event.metadata['foo'] = "bar"
document = Document('plaintext', text='Some text.')
event.add_document(document)
one = label(start_index=0, end_index=5, x=10)
two = label(start_index=6, end_index=10, x=15)
document.add_labels('one', [one,
two])
document.add_labels('two', [label(start_index=0, end_index=25, a='b', b=one),
label(start_index=26, end_index=42, a='c', b=two)])
document.add_labels('three', [
label(start_index=0, end_index=10, foo=True),
label(start_index=11, end_index=15, foo=False)
], distinct=True)
with TemporaryFile('w+') as tf:
YamlSerializer.event_to_file(event, tf)
tf.flush()
tf.seek(0)
e = YamlSerializer.file_to_event(tf)
assert e.event_id == event.event_id
assert e.metadata['foo'] == 'bar'
d = e.documents['plaintext']
assert d.text == document.text
index_one = d.labels['one']
assert index_one == [one, two]
index_two = d.labels['two']
assert index_two == [label(start_index=0, end_index=25, a='b', b=one),
label(start_index=26, end_index=42, a='c', b=two)]
index_three = d.labels['three']
assert index_three == [label(start_index=0, end_index=10, foo=True),
label(start_index=11, end_index=15, foo=False)]
def process_document(self, document: Document, params: Dict[str, Any]):
terms_index_name = params.get('terms_index', 'umls_terms')
label_negated = document.get_labeler('negated')
terms = document.get_label_index(terms_index_name)
triggers = document.labels['negation_triggers']
deps = document.get_label_index('dependencies')
upos_tags = document.get_label_index('upos_tags')
with label_negated:
for sentence in document.get_label_index('sentences'):
sentence_terms = terms.inside(sentence)
sentence_triggers = triggers.inside(sentence)
if len(sentence_triggers) > 0:
negations, _ = self.negex.check_sentence(
sentence_terms, sentence_triggers, deps, upos_tags)
for start_index, end_index in negations:
label_negated(start_index, end_index)
def process_document(self, document: Document, params: Dict[str, Any]):
terms_index_name = params.get('terms_index', 'umls_terms')
label_negated = document.get_labeler('negated')
label_trigger = document.get_labeler('negation_trigger')
terms = document.get_label_index(terms_index_name)
with label_negated, label_trigger:
for sentence in document.get_label_index('sentences'):
sentence_terms = [(t.start_index - sentence.start_index,
t.end_index - sentence.start_index)
for t in terms.inside(sentence)]
negations, triggers = self.negex.check_sentence(
sentence.text, sentence_terms)
for start_index, end_index in negations:
label_negated(sentence.start_index + start_index,
sentence.start_index + end_index)
for start_index, end_index in triggers:
label_trigger(sentence.start_index + start_index,
sentence.start_index + end_index)
def test_copy_document():
e = Event()
doc = Document(document_name='first',
text='The quick brown fox jumped over the lazy dog.')
e.add_document(doc)
with doc.get_labeler('some_index') as label:
label(0, 3, word='The')
label(4, 9, word='quick')
label(10, 15, word='brown')
processor = CopyDocument('first', 'second')
processor.process(e, {})
second = e.documents['second']
assert second is not None
assert second.labels['some_index'] == [
GenericLabel(0, 3, word='The'),
GenericLabel(4, 9, word='quick'),
GenericLabel(10, 15, word='brown')
]
def test_add_labels_distinct(mocker):
client = mocker.Mock(EventsClient)
event = Event(event_id='1', client=client)
document = Document(document_name='plaintext',
text='The quick brown fox jumped over the lazy dog.',
event=event)
labels = [
GenericLabel(0, 10, document=document, x=1),
GenericLabel(11, 15, document=document, x=2),
GenericLabel(16, 20, document=document, x=3)
]
l2 = document.add_labels('index', labels, distinct=True)
client.add_labels.assert_called_with(event_id='1',
document_name='plaintext',
index_name='index',
labels=labels,
adapter=mocker.ANY)
assert l2 == labels
assert l2.distinct
def process_document(self, document: Document, params: Dict[str, Any]):
sentences = document.labels['sentences']
sentences_text = []
for sentence in sentences:
sentences_text.append(sentence.text)
stanza_doc = self.nlp(sentences_text)
all_deps = []
all_upos_tags = []
for stanza_sentence, sentence in zip(stanza_doc.sentences, sentences):
sentence_deps, sentence_upos_tags = stanza_deps_and_upos_tags(
sentence, stanza_sentence)
all_deps.extend(sentence_deps)
all_upos_tags.extend(sentence_upos_tags)
document.add_labels('dependencies', all_deps)
document.add_labels('upos_tags', all_upos_tags)
def build_doc(self, event):
text = ' '.join(self.text)
d = Document(document_name='plaintext', text=text)
event.add_document(d)
d.add_labels('pos_tags', self.tags, distinct=True)
d.add_labels('sentences', self.sentences, distinct=True)
return d
def test_labeler_distinct(mocker):
client = mocker.Mock(EventsClient)
event = Event(event_id='1', client=client)
document = Document(document_name='plaintext',
text='The quick brown fox jumped over the lazy dog.',
event=event)
with document.get_labeler('index', distinct=True) as add_generic_label:
add_generic_label(0, 10, x=1)
add_generic_label(11, 15, x=2)
add_generic_label(16, 20, x=3)
labels = [
GenericLabel(0, 10, document=document, x=1),
GenericLabel(11, 15, document=document, x=2),
GenericLabel(16, 20, document=document, x=3)
]
label_adapter = DistinctGenericLabelAdapter
client.add_labels.assert_called_with(event_id='1',
document_name='plaintext',
index_name='index',
labels=labels,
adapter=label_adapter)
assert document.get_label_index('index') == labels
def test_add_labels_not_distinct(mocker):
client = mocker.Mock(EventsClient)
client.get_local_instance.return_value = client
client.get_label_index_info.return_value = []
event = Event(event_id='1', client=client)
document = Document(document_name='plaintext',
text='The quick brown fox jumped over the lazy dog.',
event=event)
labels = [
GenericLabel(0, 10, document=document, x=1),
GenericLabel(11, 15, document=document, x=2),
GenericLabel(16, 20, document=document, x=3)
]
document.add_labels('index', labels)
client.add_labels.assert_called_with(event_id='1',
document_name='plaintext',
index_name='index',
labels=labels,
adapter=mocker.ANY)
l2 = document.labels['index']
assert l2 == labels
assert not l2.distinct
def test_labeler_distinct(mocker):
client = mocker.Mock(EventsClient)
client.get_local_instance.return_value = client
client.get_label_index_info.return_value = []
event = Event(event_id='1', client=client)
document = Document(document_name='plaintext',
text='The quick brown fox jumped over the lazy dog.',
event=event)
with document.get_labeler('index', distinct=True) as add_generic_label:
add_generic_label(0, 10, x=1)
add_generic_label(11, 15, x=2)
add_generic_label(16, 20, x=3)
labels = [
GenericLabel(0, 10, document=document, x=1),
GenericLabel(11, 15, document=document, x=2),
GenericLabel(16, 20, document=document, x=3)
]
label_adapter = DISTINCT_GENERIC_ADAPTER
client.add_labels.assert_called_with(event_id='1',
document_name='plaintext',
index_name='index',
labels=labels,
adapter=label_adapter)
assert document.labels['index'] == labels
def main(args=None):
parser = ArgumentParser()
parser.add_argument('input',
metavar='INPUT_FOLDER',
help='A folder containing PTB formatted documents.')
parser.add_argument('--glob', metavar='GLOB', default='*.mrg')
parser.add_argument('--source-name',
metavar='DOCUMENT_NAME',
default='source',
help='What document to dump the PTB text into.')
parser.add_argument(
'--target-name',
metavar='DOCUMENT_NAME',
default='plaintext',
help='What document to the plaintext and annotations into.')
parser.add_argument('--events',
metavar='EVENTS',
default=None,
help='The address of the events service.')
parser.add_argument('--ptb-reader',
metavar='READER',
default=None,
help='The address of the PTB Reader.')
parser.add_argument('--tnt-trainer',
metavar='TRAINER',
default=None,
help='The address of the TnT trainer.')
args = parser.parse_args(args)
with EventsClient(address=args.events) as client, Pipeline(
RemoteProcessor('ptb-reader',
address=args.ptb_reader,
params={
'source_document_name': args.source_name,
'target_document_name': args.target_name
}),
RemoteProcessor('biomedicus-tnt-trainer',
address=args.tnt_trainer,
params={'document_name':
args.target_name})) as pipeline:
for f in Path(args.input).rglob(args.glob):
print('Reading:', f)
with f.open('r') as r:
text = r.read()
with Event(event_id=f.name, client=client) as event:
d = Document(args.source_name, text=text)
event.add_document(d)
pipeline.run(event)
def process_document(self,
document: mtap.Document,
params: Dict[str, Any]) -> Optional[Dict[str, Any]]:
if params['do_work']:
with self.started_stopwatch('fetch_time'):
text = document.text
a_count = text.count('a')
b_count = text.count('b')
with document.get_labeler('mtap.examples.letter_counts') as label_letter_count:
label_letter_count(start_index=0, end_index=len(document.text), letter='a',
count=a_count)
label_letter_count(start_index=0, end_index=len(document.text), letter='b',
count=b_count)
return {'answer': 42}
def main(args=None):
parser = ArgumentParser()
parser.add_argument('input',
metavar='INPUT_DIR',
help='A folder containing PTB formatted documents.')
parser.add_argument('output',
metavar='OUTPUT_DIR',
help='A folder to write the json files to.')
parser.add_argument('--glob', metavar='GLOB', default='*.mrg')
parser.add_argument('--events',
metavar='EVENTS',
default=None,
help='The address of the events service.')
parser.add_argument('--ptb-reader',
metavar='READER',
default=None,
help='The address of the PTB Reader.')
args = parser.parse_args(args)
with EventsClient(address=args.events) as client, Pipeline(
RemoteProcessor('ptb-reader',
address=args.ptb_reader,
params={
'source_document_name': 'source',
'target_document_name': 'gold',
'pos_tags_index': 'gold_tags'
}),
LocalProcessor(SerializationProcessor(JsonSerializer,
output_dir=args.output),
component_id='serializer',
client=client)) as pipeline:
for f in Path(args.input).rglob(args.glob):
print('Reading:', f)
with f.open('r') as r:
text = r.read()
with Event(event_id=f.name, client=client) as event:
d = Document('source', text=text)
event.add_document(d)
pipeline.run(event)
def test_sentences_unknown_character(bi_lstm_model):
document = Document('plaintext',
text='• Sentence which contains unknown character.')
bi_lstm_model.process_document(document, {})
assert document.get_label_index('sentences') == [GenericLabel(2, 44)]
# Copyright 2019 Regents of the University of Minnesota.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Hello world tutorial pipeline."""
import sys
if __name__ == '__main__':
from mtap import Document, Event, EventsClient, Pipeline, RemoteProcessor
with EventsClient(address=sys.argv[1]) as client, \
Pipeline(
RemoteProcessor(processor_id='hello', address=sys.argv[2])
) as pipeline:
with Event(event_id='1', client=client) as event:
document = Document(document_name='name', text='YOUR NAME')
event.add_document(document)
pipeline.run(document)
index = document.get_label_index('hello')
for label in index:
print(label.response)
def test_text_from_document():
d = Document('plaintext', text='This is text.')
assert GenericLabel(5, 7, document=d).text == 'is'
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import pytest
from mtap import GenericLabel, Location, Document
from mtap.data._label_indices import presorted_label_index, label_index
document = Document('plaintext', text='blah')
@pytest.fixture
def tested():
return presorted_label_index([
GenericLabel(0, 5, document=document, i=0),
GenericLabel(0, 7, document=document, i=1),
GenericLabel(2, 6, document=document, i=2),
GenericLabel(6, 7, document=document, i=3),
GenericLabel(6, 8, document=document, i=4),
GenericLabel(9, 10, document=document, i=5),
GenericLabel(9, 13, document=document, i=6),
GenericLabel(9, 13, document=document, i=7),
])
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import pytest
from mtap import GenericLabel, Document
document = Document('plaintext', text='foo bar')
def test_get_repr():
label = GenericLabel(0, 20, document=document, a="x", y=20, z=20.0)
rep = repr(label)
assert rep.startswith("GenericLabel(0, 20, ")
assert 'a="x"' in rep or "a='x'" in rep
assert 'y=20' in rep
assert 'z=20.0' in rep
def test_get_attr():
label = GenericLabel(0, 20, document=document, a="x", y=20, z=20.0)
def process_document(self, document: Document, params: Dict[str, Any]):
with document.get_labeler('sentences', distinct=True) as add_sentence:
for start, end in predict_text(self.model, self.input_mapper,
document.text, self.device):
add_sentence(start, end)
def process_document(self, document: Document, params: Dict[str, Any]):
text = document.text
result = self.pool.apply(predict_sentences_async, args=(text, ))
with document.get_labeler('sentences', distinct=True) as add_sentence:
for start, end in result:
add_sentence(start, end)
def process_document(self, document: Document, params: Dict[str, Any]):
with document.get_labeler('sentences') as sentence_labeler:
for start, end in get_sentences(document.text):
sentence_labeler(start, end)
def test_yml_serializer():
event = Event(event_id='1')
event.metadata['foo'] = "bar"
document = Document('plaintext', text='Some text.')
event.add_document(document)
document.add_labels('one', [
label(start_index=0, end_index=5, x=10),
label(start_index=6, end_index=10, x=15)
])
document.add_labels('two', [
label(start_index=0, end_index=25, a='b'),
label(start_index=26, end_index=42, a='c')
])
document.add_labels('three', [
label(start_index=0, end_index=10, foo=True),
label(start_index=11, end_index=15, foo=False)
],
distinct=True)
with TemporaryFile('w+') as tf:
YamlSerializer.event_to_file(event, tf)
tf.flush()
tf.seek(0)
o = yaml.load(tf, Loader=Loader)
assert o['event_id'] == '1'
assert o['metadata']['foo'] == 'bar'
d = o['documents']['plaintext']
assert d['text'] == 'Some text.'
assert len(d['label_indices']) == 3
assert d['label_indices']['one'] == {
'json_labels': [{
'start_index': 0,
'end_index': 5,
'x': 10
}, {
'start_index': 6,
'end_index': 10,
'x': 15
}],
'distinct':
False
}
assert d['label_indices']['two'] == {
'json_labels': [{
'start_index': 0,
'end_index': 25,
'a': 'b'
}, {
'start_index': 26,
'end_index': 42,
'a': 'c'
}],
'distinct':
False
}
assert d['label_indices']['three'] == {
'json_labels': [{
'start_index': 0,
'end_index': 10,
'foo': True
}, {
'start_index': 11,
'end_index': 15,
'foo': False
}],
'distinct':
True
}
