def make_uem_segments(uem_path: Pathlike,
recording: Recording) -> List[SupervisionSegment]:
lines = uem_path.read_text().splitlines()
return [
SupervisionSegment(
id=
f"{recording.id}-{int(100*float(start)):06d}-{int(100*float(end)):06d}",
recording_id=recording.id,
start=float(start),
duration=round(float(end) - float(start), ndigits=8),
) for _, _, start, end in map(str.split, lines)
]
def make_rttm_segments(rttm_path: Pathlike, recording: Recording,
metadata: Dict) -> List[SupervisionSegment]:
lines = rttm_path.read_text().splitlines()
return [
SupervisionSegment(
id=
f"{recording.id}-{speaker}-{int(100*float(start)):06d}-{int(100*(float(start)+float(duration))):06d}",
recording_id=recording.id,
start=float(start),
duration=float(duration),
speaker=speaker,
language=metadata["lang"],
custom=metadata,
) for _, _, channel, start, duration, _, _, speaker, _, _ in map(
str.split, lines)
]