def test_first_outermost_dus():
edu1 = FakeEDU('e1', span=(1, 2))
edu2 = FakeEDU('e2', span=(1, 3))
edu3 = FakeEDU('e3', span=(2, 3))
edu4 = FakeEDU('e4', span=(4, 5))
edu5 = FakeEDU('e5', span=(6, 8))
rel1 = FakeRelInst('r-e1-e2', edu1, edu2)
rel2 = FakeRelInst('r-e2-e3', edu2, edu3)
rel3 = FakeRelInst('r-e4-e5', edu4, edu5)
cdu1 = FakeCDU('c1', [edu1, edu2, edu3])
cdu2 = FakeCDU('c2', [edu4, edu5])
cdu3 = FakeCDU('c3', [cdu1, cdu2])
doc = FakeDocument([edu1, edu2, edu3, edu4, edu5],
[rel1, rel2, rel3],
[cdu1, cdu2, cdu3])
k = corpus.FileId('moo', None, None, None)
gr = stac_gr.Graph.from_doc({k: doc}, k)
ids = nodeform_graph_ids(gr)
got = gr.first_outermost_dus()
expected = ['c3', 'c1', 'e2', 'e1', 'e3', 'c2', 'e4', 'e5']
assert got == [ids[x] for x in expected]
def test_first_outermost_dus_simple():
edu1 = FakeEDU('e1', span=(1, 2))
edu2 = FakeEDU('e2', span=(1, 3))
edu3 = FakeEDU('e3', span=(2, 3))
rel1 = FakeRelInst('r-e1-e2', edu1, edu2)
rel2 = FakeRelInst('r-e2-e3', edu2, edu3)
doc = FakeDocument([edu1, edu2, edu3], [rel1, rel2], [])
k = corpus.FileId('moo', None, None, None)
gr = stac_gr.Graph.from_doc({k: doc}, k)
ids = nodeform_graph_ids(gr)
assert gr.first_outermost_dus() == [ids[x] for x in ['e2', 'e1', 'e3']]
def from_corenlp_output_filename(f):
"""
Return a tuple of FileId and turn id.
This is entirely by convention we established when calling corenlp
of course
"""
prefix = os.path.basename(f)
prefix = os.path.splitext(prefix)[0]
prefix = os.path.splitext(prefix)[0]
parts = prefix.split('_')
file_id = corpus.FileId(doc=parts[0],
subdoc=parts[1] if len(parts) > 1 else None,
stage='unannotated',
annotator=None)
turn_id = parts[-1] if len(parts) == 3 else None
return file_id, turn_id