OldToNewTranscriptions.py

class OldToNewTranscriptions(Job):
  def __init__(self, input_txt: Path):
    super(OldToNewTranscriptions, self).__init__()
    self.input_txt = input_txt
    self.output_txt = self.output_path("output.txt.gz")

  def run(self):
    import re
    from recipe.utils import generic_open
    input_transcriptions = eval(generic_open(self.input_txt.get_path()).read())
    print("Create file:", self.output_txt.get_path())
    count = 0
    with generic_open(self.output_txt.get_path(), "w") as out:
      out.write("{\n")
      for seq_tag, txt in sorted(input_transcriptions.items()):
        # old seq tag 'dev-other-116-288045-0000'
        # new seq tag 'LibriSpeech/dev-other/3663/172528/3663-172528-0000.flac.ogg'
        assert isinstance(txt, str)
        m = re.match(r"([a-z\-]+)-([0-9]+)-([0-9]+)-([0-9]+)", seq_tag)
        seq_tag = f"LibriSpeech/{m.group(1)}/{m.group(2)}/{m.group(3)}/{m.group(2)}-{m.group(3)}-{m.group(4)}.flac.ogg"
        out.write("%r: %r,\n" % (seq_tag, txt))
        count += 1
      out.write("}\n")
    print("Num seqs:", count)

  def tasks(self):
    yield Task('run', rqmt={'mem': 1, 'time': 10}, mini_task=True)