def test_online_coref(self):
# Test text, sents, tokens input
# utterances = [
# {'speaker_id': 1, 'text': 'I read an article today. It is about US politics.'},
# {'speaker_id': 2, 'tokens': [['What', 'does', 'it', 'say', 'about', 'US', 'politics', '?']]}, # Tokens
# {'speaker_id': 1, 'text': 'It talks about the US presidential election.'},
# {'speaker_id': 2, 'text': ['I am interested to hear.', 'Can you elaborate more?']}, # Sents
# {'speaker_id': 1, 'text': 'Sure! The presidential election is indeed interesting.'}
# ]
utterances = [{
'speaker_id': 1,
'tokens': [['I', 'read', 'an', 'article', 'today', '.']]
}, {
'speaker_id':
2,
'tokens':
[['Can', 'you', 'tell', 'me', 'what', 'it', 'is', 'about', '?']]
}]
context = None
tokens_to_date = []
for turn, uttr in enumerate(utterances):
if 'tokens' in uttr:
input_doc = Input(
tokens=uttr['tokens'],
speaker_ids=uttr['speaker_id'],
coref_context=self.convert_output_to_context(context),
models=['ocr'])
else:
input_doc = Input(
text=uttr['text'],
speaker_ids=uttr['speaker_id'],
coref_context=self.convert_output_to_context(context),
models=['ocr'])
output_doc = en_services.online_coref.predict_sequentially(
input_doc, check_sanitization=True)
print(json.dumps(output_doc))
context = output_doc['ocr']
# Print cluster text
tokens_to_date += flatten(input_doc.tokens)
for cluster in output_doc['ocr']['clusters']:
for i in range(len(cluster)):
m1, m2 = tuple(cluster[i])
cluster[i] = (m1, m2, ' '.join(tokens_to_date[m1:m2]))
print(f'cluster text: {output_doc["ocr"]["clusters"]}')
print()
def main():
if len(sys.argv) == 1:
sys.argv.append('--help')
arg_parser = argparse.ArgumentParser(
description='ELIT-{}'.format(__version__))
task_parser = arg_parser.add_subparsers(dest="task",
help='which task to perform?')
parse_parser = task_parser.add_parser(
name='parse', help='interactive parse per document')
server_parser = task_parser.add_parser(
name='serve',
help='start http server',
description='A http server for ELIT')
server_parser.add_argument('--port', type=int, default=8000)
server_parser.add_argument('--workers',
type=int,
default=1,
help='number of workers')
args = arg_parser.parse_args()
if args.task == 'parse':
from elit.server.en_parser import service_parser
for line in sys.stdin:
line = line.strip()
doc = service_parser.parse([Input(text=line)])[0]
print(doc)
elif args.task == 'serve':
from elit.server import server
server.run(port=args.port, workers=args.workers)
def test_sents_input(self):
text = [
"Emory NLP is a research lab in Atlanta, GA.",
"It is founded by Jinho D. Choi in 2014.",
'Dr. Choi is a professor at Emory University.'
]
doc = en_services.parser.parse([Input(text=text)])[0]
print(doc)
def test_doc_coref_sents(self):
text = [
"Emory NLP is a research lab in Atlanta, GA.",
"It is founded by Jinho D. Choi in 2014.",
'Dr. Choi is a professor at Emory University.'
]
input_doc = Input(text=text, models=['dcr'])
print(en_services.doc_coref.predict(input_doc))
def main():
text = [
"Emory NLP is a research lab in Atlanta, GA. "
"It is founded by Jinho D. Choi in 2014. Dr. Choi is a professor at Emory University."
]
input = Input(text=text)
input.models = ['lem']
docs = en_services.parser.parse([input])
for doc in docs:
print(doc)
# See elit.client for coreference examples
text = 'Pfizer said last week it may need the U.S. government to help it secure some components needed to ' \
'make the vaccine. While the company halved its 2020 production target due to manufacturing issues, ' \
'it said last week its manufacturing is running smoothly now. The government also has the option to ' \
'acquire up to an additional 400 million doses of the vaccine.'
input_doc = Input(text=text, models=['dcr'])
doc = service_doc_coref.predict(input_doc)
print(doc)
def test_doc_coref_tokens(self):
tokens = [[
"Emory", "NLP", "is", "a", "research", "lab", "in", "Atlanta", ",",
"GA", "."
],
[
"It", "is", "founded", "by", "Jinho", "D.", "Choi", "in",
"2014", ".", "Dr.", "Choi", "is", "a", "professor", "at",
"Emory", "University", "."
]]
input_doc = Input(tokens=tokens, models=['dcr'])
print(en_services.doc_coref.predict(input_doc))
def test_tokens_input(self):
tokens = [[
"Emory", "NLP", "is", "a", "research", "lab", "in", "Atlanta", ",",
"GA", "."
],
[
"It", "is", "founded", "by", "Jinho", "D.", "Choi", "in",
"2014", "."
],
[
"Dr.", "Choi", "is", "a", "professor", "at", "Emory",
"University", "."
]]
doc = en_services.parser.parse([Input(tokens=tokens)])[0]
print(doc)
def test_doc_coref_concurrent(self):
batch_size = 32
inputs = [Input(text=self.get_sample_text(), models=['dcr'])
] * batch_size
start_time = time.time()
docs = en_services.doc_coref.predict(inputs)
end_time = time.time()
print(
f'Concurrent doc coref time elapse for {batch_size} small documents: {end_time - start_time :.2f}s'
)
assert len(docs) == len(inputs)
print(docs[0])
print(docs[-1])
def test_tokens_input(self):
tokens = [
"yes i do what 's your job".split(),
]
doc = en_services.parser.parse([Input(tokens=tokens)])[0]
print(doc)
async def parse(text: str):
input = Input(text=text)
output: Document = await runner.process_input(input)
if not isinstance(output, Document):
raise HandlingError("Internal Server Error", code=500)
return output.to_dict()