def main(args=None):
parser = ArgumentParser()
parser.add_argument("input_directory", metavar="INPUT_DIR")
parser.add_argument("output_directory", metavar="OUTPUT_DIR")
parser.add_argument("--events")
parser.add_argument("--tagger")
parser.add_argument("--sentences")
parser.add_argument("--acronyms")
parser.add_argument("--norms")
parser.add_argument("--concepts")
args = parser.parse_args(args)
input_dir = Path(args.input_directory)
with EventsClient(address=args.events) as client, Pipeline(
RemoteProcessor('biomedicus-sentences', address=args.sentences),
RemoteProcessor('biomedicus-tnt-tagger', address=args.tagger),
RemoteProcessor('biomedicus-acronyms', address=args.acronyms),
RemoteProcessor('biomedicus-concepts', address=args.concepts),
LocalProcessor(SerializationProcessor(
JsonSerializer, output_dir=args.output_directory),
component_id='serialize',
client=client)) as pipeline:
for path in input_dir.glob("**/*.txt"):
print("READING FILE:", str(path))
with path.open('r') as f:
contents = f.read()
with Event(event_id=path.stem, client=client) as event:
document = event.create_document("plaintext", text=contents)
pipeline.run(document)
pipeline.print_times()
def main(args=None):
parser = ArgumentParser()
parser.add_argument('--events-service')
parser.add_argument('--sentences-service')
parser.add_argument('--dependencies-service')
parser.add_argument('input_file')
conf = parser.parse_args(args)
with EventsClient(address=conf.events_service) as client, \
Pipeline(
RemoteProcessor('biomedicus-sentences', address=conf.sentences_service),
RemoteProcessor('biomedicus-dependencies', address=conf.dependencies_service)
) as pipeline:
with open(conf.input_file, 'r') as in_f:
txt = in_f.read()
with Event(event_id=Path(conf.input_file).name,
client=client) as event:
document = event.create_document('plaintext', txt)
pipeline.run(document)
for sentence in document.labels['sentences']:
print(sentence.text)
print('\n')
for dependency in document.labels['dependencies'].inside(
sentence):
print((dependency.text, dependency.deprel,
dependency.head.text
if dependency.head is not None else 'ROOT'))
print('\n')
def test_deepen_performance(events_service, negex_triggers_service,
dependencies_service, deepen_negation_service,
test_results):
input_dir = Path(
os.environ['BIOMEDICUS_TEST_DATA']) / 'negation' / 'i2b2_2010'
confusion = metrics.FirstTokenConfusion()
metrics_processor = metrics.Metrics(confusion,
tested='negated',
target='i2b2concepts',
target_filter=is_negated)
with EventsClient(address=events_service) as client, Pipeline(
RemoteProcessor(processor_id='biomedicus-negex-triggers',
address=negex_triggers_service),
RemoteProcessor(processor_id='biomedicus-selective-dependencies',
address=dependencies_service,
params={'terms_index': 'i2b2concepts'}),
RemoteProcessor('biomedicus-deepen',
address=deepen_negation_service,
params={'terms_index': 'i2b2concepts'}),
LocalProcessor(metrics_processor, component_id='metrics'),
events_client=client) as pipeline:
for test_file in input_dir.glob('**/*.pickle'):
with PickleSerializer.file_to_event(test_file,
client=client) as event:
document = event.documents['plaintext']
results = pipeline.run(document)
print('F1 for event - "{}": {:0.3f} - elapsed: {}'.format(
event.event_id,
results.component_result(
'metrics').result_dict['first_token_confusion']['f1'],
results.component_result(
'biomedicus-selective-dependencies').
timing_info['process_method']))
print('Overall Precision:', confusion.precision)
print('Overall Recall:', confusion.recall)
print('Overall F1:', confusion.f1)
pipeline.print_times()
timing_info = pipeline.processor_timer_stats(
'biomedicus-deepen').timing_info
timing_info_parse = pipeline.processor_timer_stats(
'biomedicus-selective-dependencies').timing_info
test_results['biomedicus-deepen'] = {
'Gold Standard':
"2010 i2b2-VA",
'Precision':
confusion.precision,
'Recall':
confusion.recall,
'F1':
confusion.f1,
'Per-Document Mean Remote Call Duration':
str(timing_info['remote_call'].mean +
timing_info_parse['remote_call'].mean),
'Per-Document Mean Process Method Duration':
str(timing_info['process_method'].mean +
timing_info_parse['process_method'].mean)
}
def main(args=None):
parser = ArgumentParser()
parser.add_argument("input_directory", metavar="INPUT_DIR")
parser.add_argument("concepts_csv", metavar="PATH_TO_CONCEPTS_CSV")
parser.add_argument("output_directory", metavar="OUTPUT_DIR")
parser.add_argument("--sentences")
parser.add_argument("--tagger")
parser.add_argument("--acronyms")
parser.add_argument("--events")
ns = parser.parse_args(args)
print('Reading concepts csv...')
concepts = {}
with open(ns.concepts_csv, 'r') as f:
for line in f.readlines():
splits = line.split(',')
end = splits[0]
start = splits[1]
cui = splits[5]
identifier = splits[6]
try:
v = concepts[identifier]
except KeyError:
v = []
concepts[identifier] = v
v.append((start, end, cui))
print('Reading mipacq source files...')
with EventsClient(address=ns.events) as client, \
Pipeline(
RemoteProcessor('biomedicus-sentences', address=ns.sentences),
RemoteProcessor('biomedicus-tnt-tagger', address=ns.tagger),
RemoteProcessor('biomedicus-acronyms', address=ns.acronyms),
LocalProcessor(SerializationProcessor(PickleSerializer,
output_dir=ns.output_directory),
component_id='serialize',
client=client)
) as pipeline:
for path in Path(ns.input_directory).glob('**/*.source'):
identifier = path.stem.split('-')[0]
try:
doc_concepts = concepts[identifier]
except KeyError:
continue
with Event(event_id=identifier, client=client) as event:
with path.open('r') as f:
text = f.read()
document = event.create_document('plaintext', text)
with document.get_labeler('gold_concepts') as label_concept:
for start, end, cui in doc_concepts:
label_concept(start, end, cui=cui)
pipeline.run(document)
def run_default_pipeline(config: Namespace):
with DefaultPipeline(
conf_path=config.config,
output_directory=config.output_directory,
events_addresses=config.events_addresses,
serializer=config.serializer,
include_label_text=config.include_label_text) as default_pipeline:
if config.rtf:
rtf_processor = RemoteProcessor(
processor_id='biomedicus-rtf',
address=config.rtf_address,
params={'output_document_name': 'plaintext'})
default_pipeline.pipeline.insert(0, rtf_processor)
extension_glob = config.extension_glob or "**/*.rtf"
source = rtf_source(config.input_directory, extension_glob,
default_pipeline.pipeline.events_client)
params = {'document_name': 'plaintext'}
else:
extension_glob = config.extension_glob or "**/*.txt"
source = FilesInDirectoryProcessingSource(
default_pipeline.pipeline.events_client,
config.input_directory,
extension_glob=extension_glob)
params = None
default_pipeline.pipeline.run_multithread(source, params=params)
default_pipeline.pipeline.print_times()
def main(args=None):
parser = ArgumentParser()
parser.add_argument('input',
metavar='INPUT_FILE',
help='The input GENIA XML file.')
parser.add_argument('--events',
metavar='EVENTS',
default=None,
help='The address of the events service.')
parser.add_argument('--tnt-trainer',
metavar='TRAINER',
default=None,
help='The address of the TnT trainer.')
args = parser.parse_args(args)
etree = ElementTree.parse(args.input)
set = etree.getroot()
with EventsClient(args.events) as client, Pipeline(
RemoteProcessor('biomedicus-tnt-trainer',
address=args.tnt_trainer)) as pipeline:
for article in set.findall('article'):
id = list(article.find('articleinfo'))[0].text
with Event(id, client) as event:
db = DocumentBuilder()
for sentence in article.find('title').findall(
'sentence') + article.find('abstract').findall(
'sentence'):
db.add_sentence(sentence)
d = db.build_doc(event)
pipeline.run(d)
def test_tnt_performance(events_service, pos_tags_service, test_results):
input_dir = Path(os.environ['BIOMEDICUS_TEST_DATA']) / 'pos_tags'
accuracy = Accuracy()
with EventsClient(address=events_service) as client, Pipeline(
RemoteProcessor(processor_id='biomedicus-tnt-tagger',
address=pos_tags_service,
params={'token_index': 'gold_tags'}),
LocalProcessor(Metrics(accuracy,
tested='pos_tags',
target='gold_tags'),
component_id='metrics'),
events_client=client) as pipeline:
for test_file in input_dir.glob('**/*.pickle'):
event = PickleSerializer.file_to_event(test_file, client=client)
with event:
document = event.documents['gold']
results = pipeline.run(document)
print(
'Accuracy for event - ', event.event_id, ':',
results.component_result(
'metrics').result_dict['accuracy'])
print('Accuracy:', accuracy.value)
pipeline.print_times()
timing_info = pipeline.processor_timer_stats(
'biomedicus-tnt-tagger').timing_info
test_results['TnT Pos Tagger'] = {
'Accuracy': accuracy.value,
'Remote Call Duration': str(timing_info['remote_call'].mean),
'Process Method Duration': str(timing_info['process_method'].mean)
}
assert accuracy.value > 0.9
def test_sentence_performance(events_service, sentences_service, test_results):
input_dir = Path(os.environ['BIOMEDICUS_TEST_DATA']) / 'sentences'
confusion = metrics.FirstTokenConfusion()
with EventsClient(address=events_service) as client, Pipeline(
RemoteProcessor(processor_id='biomedicus-sentences',
address=sentences_service),
LocalProcessor(metrics.Metrics(confusion,
tested='sentences',
target='Sentence'),
component_id='metrics',
client=client)) as pipeline:
for test_file in input_dir.glob('**/*.json'):
with JsonSerializer.file_to_event(test_file,
client=client) as event:
document = event.documents['plaintext']
results = pipeline.run(document)
print('F1 for event - "{}": {:0.3f} - elapsed: {}'.format(
event.event_id,
results[1].results['first_token_confusion']['f1'],
results[0].timing_info['process_method']))
print('Overall Precision:', confusion.precision)
print('Overall Recall:', confusion.recall)
print('Overall F1:', confusion.f1)
pipeline.print_times()
timing_info = pipeline.processor_timer_stats()[0].timing_info
test_results['Sentences'] = {
'Precision': confusion.precision,
'Recall': confusion.recall,
'F1': confusion.f1,
'Remote Call Duration': str(timing_info['remote_call'].mean),
'Process Method Duration': str(timing_info['process_method'].mean)
}
assert confusion.f1 > 0.85
def test_concepts_performance(events_service, concepts_service, test_results):
input_dir = Path(os.environ['BIOMEDICUS_TEST_DATA']) / 'concepts'
recall = Accuracy(name='recall', mode='any', fields=['cui'])
precision = Accuracy(name='precision', mode='any', fields=['cui'])
with EventsClient(address=events_service) as client, \
Pipeline(
RemoteProcessor(processor_id='biomedicus-concepts', address=concepts_service),
LocalProcessor(Metrics(recall, tested='umls_concepts', target='gold_concepts'),
component_id='metrics'),
LocalProcessor(Metrics(precision, tested='gold_concepts', target='umls_concepts'),
component_id='metrics_reverse'),
events_client=client
) as pipeline:
for test_file in input_dir.glob('**/*.pickle'):
with PickleSerializer.file_to_event(test_file, client=client) as event:
document = event.documents['plaintext']
pipeline.run(document)
print('Precision:', precision.value)
print('Recall:', recall.value)
timing_info = pipeline.processor_timer_stats('biomedicus-concepts').timing_info
test_results['Concepts'] = {
'Precision': precision.value,
'Recall': recall.value,
'Remote Call Duration': str(timing_info['remote_call'].mean),
'Process Method Duration': str(timing_info['process_method'].mean)
}
assert recall.value > 0.6
def test_java_references(python_events, java_references_processor):
with EventsClient(address=python_events) as client, Pipeline(
RemoteProcessor('mtap-java-reference-labels-example-processor',
address=java_references_processor)
) as pipeline:
with Event(event_id='1', client=client) as event:
document = event.create_document('plaintext', 'abcd')
pipeline.run(document)
references = document.labels['references']
assert references[0].a == GenericLabel(0, 1)
assert references[0].b == GenericLabel(1, 2)
assert references[1].a == GenericLabel(2, 3)
assert references[1].b == GenericLabel(3, 4)
map_references = document.labels['map_references']
assert map_references[0].ref == {
'a': GenericLabel(0, 1),
'b': GenericLabel(1, 2),
'c': GenericLabel(2, 3),
'd': GenericLabel(3, 4)
}
list_references = document.labels['list_references']
assert list_references[0].ref == [GenericLabel(0, 1), GenericLabel(1, 2)]
assert list_references[1].ref == [GenericLabel(2, 3), GenericLabel(3, 4)]
def test_dependencies(events_service, dependencies_service, test_results):
test_dir = Path(os.environ['BIOMEDICUS_TEST_DATA']) / 'dependencies'
uas = Accuracy('UAS', equivalence_test=uas_equal)
las = Accuracy('LAS', equivalence_test=las_equal)
with EventsClient(address=events_service) as client, \
Pipeline(
RemoteProcessor(processor_id='biomedicus-dependencies',
address=dependencies_service),
LocalProcessor(Metrics(uas, las, tested='dependencies', target='gold_dependencies'),
component_id='accuracy', client=client)
) as pipeline:
for test_file in test_dir.glob('**/*.pickle'):
with PickleSerializer.file_to_event(test_file,
client=client) as event:
document = event.documents['plaintext']
results = pipeline.run(document)
accuracy_dict = results.component_result(
'accuracy').result_dict
print('Results for document: UAS: {}. LAS: {}.'.format(
accuracy_dict['UAS'], accuracy_dict['LAS']))
print('UAS:', uas.value)
print('LAS:', las.value)
timing_info = pipeline.processor_timer_stats(
'biomedicus-dependencies').timing_info
test_results['biomedicus-dependencies'] = {
'UAS': uas.value,
'LAS': las.value,
'Corpus': "MiPACQ converted to UD from PTB test set",
'Remote Call Duration': str(timing_info['remote_call'].mean),
'Process Method Duration': str(timing_info['process_method'].mean)
}
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 test_acronyms_performance(events_service, acronyms_service, test_results):
input_dir = Path(os.environ['BIOMEDICUS_PHI_TEST_DATA']) / 'acronyms'
top_score_accuracy = Accuracy(name='top_score_accuracy',
fields=['expansion'])
any_accuracy = Accuracy(name='any_accuracy',
mode='any',
fields=['expansion'])
detection_recall = Accuracy(name='detection_recall',
mode='location',
fields=['expansion'])
detection_precision = Accuracy(name='detection_precision',
mode='location',
fields=['expansion'])
with EventsClient(address=events_service) as client, Pipeline(
RemoteProcessor(processor_id='biomedicus-acronyms',
address=acronyms_service),
LocalProcessor(Metrics(top_score_accuracy,
detection_recall,
tested='acronyms',
target='gold_acronyms'),
component_id='top_score_metrics',
client=client),
LocalProcessor(Metrics(detection_precision,
tested='gold_acronyms',
target='acronyms'),
component_id='top_score_reverse',
client=client),
LocalProcessor(Metrics(any_accuracy,
tested='all_acronym_senses',
target='gold_acronyms'),
component_id='all_senses_metrics',
client=client)) as pipeline:
for test_file in input_dir.glob('**/*.json'):
with JsonSerializer.file_to_event(test_file,
client=client) as event:
document = event.documents['plaintext']
pipeline.run(document)
print('Top Sense Accuracy:', top_score_accuracy.value)
print('Any Sense Accuracy:', any_accuracy.value)
print('Detection Recall:', detection_recall.value)
print('Detection Precision:', detection_precision.value)
pipeline.print_times()
timing_info = pipeline.processor_timer_stats(
'biomedicus-acronyms').timing_info
test_results['acronyms'] = {
'Top sense accuracy': top_score_accuracy.value,
'Any sense accuracy': any_accuracy.value,
'Detection Recall': detection_recall.value,
'Detection Precision': detection_precision.value,
'Remote Call Duration': str(timing_info['remote_call'].mean),
'Process Method Duration': str(timing_info['process_method'].mean)
}
assert top_score_accuracy.value > 0.4
assert any_accuracy.value > 0.4
assert detection_recall.value > 0.65
def test_disc_pipeline(disc_python_events, disc_python_processor, disc_java_processor):
with EventsClient(address=disc_python_events) as client, mtap.Pipeline(
RemoteProcessor('mtap-example-processor-python', address='localhost:50501',
params={'do_work': True}),
RemoteProcessor('mtap-example-processor-java', address='localhost:50502',
params={'do_work': True})
) as pipeline:
with Event(event_id='1', client=client) as event:
event.metadata['a'] = 'b'
document = event.create_document('plaintext', PHASERS)
pipeline.run(document)
letter_counts = document.get_label_index('mtap.examples.letter_counts')
a_counts = letter_counts[0]
assert a_counts.count == 23
b_counts = letter_counts[1]
assert b_counts.count == 6
pipeline.print_times()
thes = document.get_label_index("mtap.examples.word_occurrences")
assert thes[0].start_index == 121
assert thes[0].end_index == 124
def __init__(self,
conf: PipelineConf,
*,
events_client: EventsClient = None):
conf.populate_addresses()
if events_client is not None:
self.close_client = False
self.events_client = events_client
elif conf.events_address is not None:
self.close_client = True
self.events_client = EventsClient(address=conf.events_address)
else:
raise ValueError("Events client or address not specified.")
pipeline = [(conf.sentences_id, conf.sentences_address),
(conf.section_headers_id, conf.section_headers_address),
(conf.tagger_id, conf.tagger_address),
(conf.acronyms_id, conf.acronyms_address),
(conf.concepts_id, conf.concepts_address),
(conf.negation_id, conf.negation_address),
(conf.selective_dependencies_id,
conf.selective_dependencies_address),
(conf.deepen_id, conf.deepen_address)]
if conf.use_discovery:
self.pipeline = Pipeline(
*[RemoteProcessor(identifier) for identifier, _ in pipeline])
else:
self.pipeline = Pipeline(*[
RemoteProcessor(identifier, address=addr)
for identifier, addr in pipeline
])
if conf.serializer is not None:
serialization_proc = SerializationProcessor(
get_serializer(conf.serializer),
conf.output_directory,
include_label_text=conf.include_label_text)
ser_comp = LocalProcessor(serialization_proc,
component_id='serializer',
client=self.events_client)
self.pipeline.append(ser_comp)
def main(args=None):
parser = ArgumentParser()
parser.add_argument("input_directory", metavar="INPUT_DIR")
parser.add_argument("output_directory", metavar="OUTPUT_DIR")
parser.add_argument("--events")
parser.add_argument("--rtf")
parser.add_argument("--tagger")
parser.add_argument("--acronyms")
parser.add_argument("--sentences")
args = parser.parse_args(args)
input_dir = Path(args.input_directory)
with EventsClient(address=args.events) as client, Pipeline(
RemoteProcessor('rtf-processor',
address=args.rtf,
params={
'binary_data_name': 'rtf',
'output_document_name': 'plaintext'
}),
RemoteProcessor('sentences',
address=args.sentences,
params={'document_name': 'plaintext'}),
RemoteProcessor('tnt-tagger',
address=args.tagger,
params={'document_name': 'plaintext'}),
RemoteProcessor('acronyms', address=args.acronyms),
LocalProcessor(SerializationProcessor(
JsonSerializer, output_dir=args.output_directory),
component_id='serialize',
client=client)) as pipeline:
for path in input_dir.glob("**/*.rtf"):
with path.open('rb') as f:
contents = f.read()
with Event(event_id=path.stem, client=client) as event:
event.binaries['rtf'] = contents
pipeline.run(event)
pipeline.print_times()
def __init__(self,
conf: PipelineConf,
*,
events_client: EventsClient = None):
conf.populate_addresses()
if events_client is not None:
self.close_client = False
self.events_client = events_client
elif conf.events_address is not None:
self.close_client = True
self.events_client = EventsClient(address=conf.events_address)
else:
raise ValueError("Events client or address not specified.")
pipeline = [(conf.sentences_id, conf.sentences_address),
(conf.tagger_id, conf.tagger_address)]
if conf.use_discovery:
self.pipeline = Pipeline(
*[RemoteProcessor(identifier) for identifier, _ in pipeline])
else:
self.pipeline = Pipeline(*[
RemoteProcessor(identifier, address=addr)
for identifier, addr in pipeline
])
def run_themes_pipeline(input_directory, annotations_directory,
output_directory):
events_address = 'localhost:50100'
with Pipeline(RemoteProcessor('biomedicus-sentences',
address='localhost:50300'),
LocalProcessor(
AttachPalliativeThemesProcessor(annotations_directory)),
LocalProcessor(CoalescePalliativeThemesProcessor(), ),
LocalProcessor(
SerializationProcessor(JsonSerializer,
output_directory)),
events_address=events_address) as pipeline:
source = FilesInDirectoryProcessingSource(pipeline.events_client,
input_directory)
pipeline.run_multithread(source, workers=8)
def test_normalization(events_service, normalization_processor):
with EventsClient(address=events_service) as client, \
Pipeline(RemoteProcessor(processor_id='biomedicus_normalizer',
address=normalization_processor)) as pipeline, \
PickleSerializer.file_to_event(Path(__file__).parent / '97_95.pickle',
client=client) as event:
document = event.documents['plaintext']
pipeline.run(document)
for norm_form in document.get_label_index('norm_forms'):
if norm_form.text == "according":
assert norm_form.norm == "accord"
if norm_form.text == "expressing":
assert norm_form.norm == "express"
if norm_form.text == "receiving":
assert norm_form.norm == "receive"
if norm_form.text == "days":
assert norm_form.norm == "day"
def main(args=None):
parser = ArgumentParser()
parser.add_argument('--events-service', default='localhost:10100')
parser.add_argument('--sentences-service', default='localhost:10102')
conf = parser.parse_args(args)
with Pipeline(
RemoteProcessor(
'biomedicus-sentences',
address=conf.sentences_service)) as pipeline, EventsClient(
address=conf.events_service) as events_client:
text = sys.stdin.read()
with Event(client=events_client) as event:
doc = event.create_document('plaintext', text)
result = pipeline.run(doc)
for sentence in doc.get_label_index('sentences'):
print('S: "', sentence.text, '"')
for k, v in result[0].timing_info.items():
print('{}: {}'.format(k, v))
def main(args=None):
parser = ArgumentParser(
description=
'Converts files from the i2b2/VA 2010 format to serialized MTAP events '
'containing the ')
parser.add_argument(
'input_directory',
type=Path,
help=
'An input directory containing a "txt" folder containing text files '
'and an "ast" folder containing the assertions in the i2b2/VA '
'pipe-delimited format.')
parser.add_argument(
'output_directory',
type=Path,
help='An output directory to write the serialized mtap events to.')
parser.add_argument('--target-document', default='plaintext')
parser.add_argument('--serializer',
default='pickle',
choices=standard_serializers.keys(),
help='The serializer to use.')
parser.add_argument('--events', help="Address of the events client.")
parser.add_argument('--tagger', help="Address of the pos tagger to use.")
conf = parser.parse_args(args)
serializer = standard_serializers[conf.serializer]
with EventsClient(address=conf.events) as client, Pipeline(
LocalProcessor(OnePerLineSentencesProcessor(),
component_id='sentences',
client=client),
RemoteProcessor('biomedicus-tnt-tagger', address=conf.tagger),
LocalProcessor(SerializationProcessor(
serializer, output_dir=conf.output_directory),
component_id='serializer',
client=client)) as pipeline:
results = pipeline.run_multithread(
events(conf.input_directory, conf.target_document, client=client))
pipeline.print_times()
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 run_sentences_pipeline(input_directory, skip_file, output_directory):
skip_documents = set(Path(skip_file).open('r').read().splitlines())
events_address = 'localhost:50100'
with Pipeline(RemoteProcessor('biomedicus-sentences',
address='localhost:50300'),
LocalProcessor(
SerializationProcessor(JsonSerializer,
output_directory)),
events_address=events_address) as pipeline:
total = sum(1 for _ in input_directory.rglob('*.txt'))
def source():
for path in input_directory.rglob('*.txt'):
relative = str(path.relative_to(input_directory))
if relative not in skip_documents:
with path.open('r') as f:
txt = f.read()
with Event(event_id=relative,
client=pipeline.events_client,
only_create_new=True) as e:
doc = e.create_document('plaintext', txt)
yield doc
pipeline.run_multithread(source(), total=total, workers=8)
# 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)
