def test_s2_spo():
deps = nlp.prepare_deps(s2)
chunks = nlp.prepare_chunks(s2)
trigger = 'causes'
s_head, s, p, o_head, o = extract_spo(deps, chunks, trigger)
assert s_head == 'virus'
assert s == 'the prototype virus' # TODO
assert p == 'causes'
assert o_head == 'gastroenteritis'
assert o == 'epidemic gastroenteritis'
def test_s1_spo():
deps = nlp.prepare_deps(s1)
chunks = nlp.prepare_chunks(s1)
trigger = 'leads to'
s_head, s, p, o_head, o = extract_spo(deps, chunks, trigger)
assert s_head == 'encapsulation'
assert s == 'The encapsulation of rifampicin'
assert p == 'leads to'
assert o_head == 'reduction'
assert o == 'a reduction of the Mycobacterium smegmatis inside macrophages'
def test_s5_spo():
deps = nlp.prepare_deps(s5)
chunks = nlp.prepare_chunks(s5)
trigger = 'cause of'
s_head, s, p, o_head, o = extract_spo(deps, chunks, trigger)
assert s_head == 'infection'
assert s == 'Chronic hepatitis virus infection'
assert p == 'cause of'
assert o_head == 'hepatitis'
assert o == 'chronic hepatitis, cirrhosis, and hepatocellular carcinoma worldwide'
def test_s4_spo():
deps = nlp.prepare_deps(s4)
chunks = nlp.prepare_chunks(s4)
trigger = 'inhibit'
s_head, s, p, o_head, o = extract_spo(deps, chunks, trigger)
assert s_head == 'ribavirin'
assert s == 'ribavirin'
assert p == 'inhibit'
assert o_head == 'replication'
assert o == 'SARS coronavirus replication in five different cell types of animal or human origin'
def test_s3_spo():
deps = nlp.prepare_deps(s3)
chunks = nlp.prepare_chunks(s3)
print(chunks)
trigger = 'cause'
s_head, s, p, o_head, o = extract_spo(deps, chunks, trigger)
assert s_head == 'exposure'
assert s == 'the exposure to ambient air pollution'
assert p == 'cause'
assert o_head == 'illnesses'
assert o == 'serious respiratory illnesses'
def main():
"""
:return:
"""
parser = argparse.ArgumentParser(description='')
parser.add_argument('-i', '--input-dir', required=True, help='...')
parser.add_argument('-o', '--output-file', required=True, help='...')
args = parser.parse_args()
dr = DataReader(args.input_dir)
it = iter(dr)
nlp = StanzaNLP()
with open(args.output_file, 'w', newline='') as csvfile:
writer = csv.writer(csvfile, delimiter='\t')
while True:
try:
pmcid, title, abstract = next(it)
# TODO just sentence splitter?
except StopIteration:
break
else:
doc = nlp.process(abstract)
for sentence in list(doc.sentences):
sent = StanzaNLP.get_sentence(sentence.words)
trigger = get_fired_trigger(sent)
if not trigger:
continue
deps = StanzaNLP.get_dependencies(sentence)
# chunking
chunks = nlp.prepare_chunks(sent)
s_head, s, p, o_head, o = extract_spo(deps, chunks, trigger)
ss = get_coordinated_nps(s)
os = get_coordinated_nps(o)
for s in ss:
for o in os:
row = [title, pmcid, f'PMC{pmcid}.nxml', s, p, o, sent]
if all(row):
writer.writerow(row)