howl/training/align/stub.py at master · castorini/howl

Repository navigation

import numpy as np

from .base import AlignedTranscription, Aligner

class StubAligner(Aligner):

"""Generate transcription with stub alignments"""

def align(self, audio) -> AlignedTranscription:

"""Generate alignment for given audio by evenly splitting the characters of transcription"""

start = 0

end = audio.audio_data.size(0) / audio.sample_rate * 1000

transcription = audio.metadata.transcription.lower()

return AlignedTranscription(

transcription=transcription, end_timestamps=np.linspace(start, end, len(transcription)).tolist()

)