def test_deterministic(self, prep_org_data):
""" Ensures loading and processing utterences from ELAN files is
deterministic.
"""
bkw_org_path = prep_org_data
utterances_1 = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
utterances_2 = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
assert utterances_1 == utterances_2
utterances_1 = [
utter for utter in utterances_1 if bkw.bkw_filter(utter)
]
utterances_2 = [
utter for utter in utterances_2 if bkw.bkw_filter(utter)
]
assert utterances_1 == utterances_2
utterances_1 = utterance.remove_duplicates(utterances_1)
utterances_2 = utterance.remove_duplicates(utterances_2)
assert utterances_1 == utterances_2
utterances_1 = [
bkw.bkw_label_segmenter.segment_labels(utter)
for utter in utterances_1
]
utterances_2 = [
bkw.bkw_label_segmenter.segment_labels(utter)
for utter in utterances_2
]
assert utterances_1 == utterances_2
utterances_1 = utterance.remove_empty_text(utterances_1)
utterances_2 = utterance.remove_empty_text(utterances_2)
assert utterances_1 == utterances_2
def test_utterances_from_dir(self, prep_org_data):
bkw_org_path = prep_org_data
utterances = elan.utterances_from_dir(bkw_org_path, ["xv"])
assert len(utterances) == 1036
assert len(utterance.remove_empty_text(utterances)) == 1035
assert len(utterance.remove_duplicates(utterances)) == 1029
assert len(
utterance.remove_duplicates(
utterance.remove_empty_text(utterances))) == 1028
utterances = elan.utterances_from_dir(bkw_org_path, ["rf"])
assert len(utterances) == 1242
assert len(utterance.remove_empty_text(utterances)) == 631
assert len(utterance.remove_duplicates(utterances)) == 1239
assert len(
utterance.remove_duplicates(
utterance.remove_empty_text(utterances))) == 631
utterances = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
assert len(utterances) == 2278
assert len(utterance.remove_empty_text(utterances)) == 1666
assert len(utterance.remove_duplicates(utterances)) == 1899
assert len(
utterance.remove_duplicates(
utterance.remove_empty_text(utterances))) == 1291
def test_poly_durations(self, prep_org_data):
bkw_org_path = prep_org_data
utterances = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
print("Total duration of utterances is {}".format(
utterance.duration(utterances)))
print("Total duration of the first utterance is {}".format(
utterance.duration(utterances[0])))
def test_explore_code_switching(self, prep_org_data):
bkw_org_path = prep_org_data
utterances = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
utterances = utterance.remove_empty_text(
utterance.remove_duplicates(utterances))
codeswitched_path = tgt_dir / "codeswitched.txt"
bkw.explore_code_switching(utterances, codeswitched_path)
def test_speaker_durations(self, prep_org_data):
bkw_org_path = prep_org_data
utterances = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
print(len(utterances))
utterances = utterance.remove_empty_text(utterances)
print(len(utterances))
utterances = utterance.remove_duplicates(utterances)
print(len(utterances))
utterances = [utter for utter in utterances if bkw.bkw_filter(utter)]
print(len(utterances))
utterances = [
utter for utter in utterances if utterance.duration(utter) < 10000
]
total = 0
fmt = "{:20}{:10}"
dur_fmt = "{:<10.3f}"
print(fmt.format("Speaker", "Duration"))
for speaker, duration in sorted(
utterance.speaker_durations(utterances),
key=lambda x: x[1],
reverse=True):
dur_mins = (duration * ureg.milliseconds).to(
ureg.minutes).magnitude
dur_str = dur_fmt.format(dur_mins)
total += dur_mins
print(fmt.format(speaker, dur_str))
print(fmt.format("Total", dur_fmt.format(total)))
def test_empty_wav(self, prep_org_data):
# Checking the origin of the empty wav.
bkw_org_path = prep_org_data
utterances = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
filtered = utterance.remove_too_short(utterances)
if filtered != utterances:
diff = set(utterances) - set(filtered)
print("set(utterances) - set(filtered): {}:\n".format(
pprint.pformat(diff)))
assert False
def test_speaker_id(self, prep_org_data):
bkw_org_path = prep_org_data
utterances = elan.utterances_from_dir(bkw_org_path, ["rf", "xv"])
no_speaker_tiers = set()
speaker_tiers = set()
speakers = set()
for utter in utterances:
tier_id = splitext(utter.prefix)[0]
if utter.speaker == None:
no_speaker_tiers.add(tier_id)
else:
speaker_tiers.add((tier_id, utter.speaker))
speakers.add(utter.speaker)
assert len(no_speaker_tiers) == 0
assert len(speakers) == NUM_SPEAKERS