File size: 1,815 Bytes
9b337f2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 |
import datasets
class AudioDataset(datasets.GeneratorBasedBuilder):
def _info(self):
return datasets.DatasetInfo(
features=datasets.Features(
{
"file_name": datasets.Value("string"),
"transcription": datasets.Value("string"),
"speaker_id": datasets.Value("int64"),
"audio": datasets.Audio(sampling_rate=16_000), # adjust the sampling rate if needed
}
),
)
def _split_generators(self, dl_manager):
data_dir = "data" # Base directory containing both metadata and audio subdirectory
return [
datasets.SplitGenerator(
name=datasets.Split.TRAIN,
gen_kwargs={"filepath": f"{data_dir}/train_metadata.csv", "audio_dir": f"{data_dir}/audio"},
),
datasets.SplitGenerator(
name=datasets.Split.TEST,
gen_kwargs={"filepath": f"{data_dir}/test_metadata.csv", "audio_dir": f"{data_dir}/audio"},
),
]
def _generate_examples(self, filepath, audio_dir):
with open(filepath, "r", encoding="utf-8") as f:
for id_, line in enumerate(f):
if id_ == 0:
continue # skip header
parts = line.strip().split(",")
file_name = parts[0]
transcription = parts[1]
speaker_id = int(parts[2])
yield id_, {
"file_name": file_name,
"transcription": transcription,
"speaker_id": speaker_id,
"audio": f"{audio_dir}/{file_name.split('/')[-1]}", # Extract the filename from the path and prepend the audio directory
}
|