corpus.py 文件源码

def save(self, corpus_csv_file: Path, use_relative_audio_file_paths: bool = True):
        import csv
        with corpus_csv_file.open('w', encoding='utf8') as opened_csv:
            writer = csv.writer(opened_csv, delimiter=',', quotechar='"', quoting=csv.QUOTE_MINIMAL)

            examples_and_phase = [(e, Phase.training) for e in self.training_examples] + \
                                 [(e, Phase.test) for e in self.test_examples]

            for e, phase in examples_and_phase:
                writer.writerow(
                    (e.id, str(e.audio_file.relative_to(
                        corpus_csv_file.parent) if use_relative_audio_file_paths else e.audio_file),
                     e.label, phase.value, e.positional_label.serialize() if e.positional_label else ""))