Datasets:
Tasks:
Automatic Speech Recognition
Formats:
webdataset
Languages:
Uzbek
Size:
10K - 100K
Tags:
audio
License:
Create stt_uz_structure.py
Browse files- stt_uz_structure.py +116 -0
stt_uz_structure.py
ADDED
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import tarfile
|
3 |
+
import csv
|
4 |
+
import datasets
|
5 |
+
from datasets.utils.py_utils import size_str
|
6 |
+
from tqdm import tqdm
|
7 |
+
|
8 |
+
|
9 |
+
class STTUzbekConfig(datasets.BuilderConfig):
|
10 |
+
"""BuilderConfig for the STT Uzbek Dataset."""
|
11 |
+
|
12 |
+
def __init__(self, **kwargs):
|
13 |
+
description = (
|
14 |
+
"Speech-to-Text dataset for the Uzbek language. "
|
15 |
+
"The dataset contains audio files stored in different folders: "
|
16 |
+
"`wavs`, `uz_other_dataset`, `uz_validated_dataset`, `uz_train_dataset`. "
|
17 |
+
"The corresponding transcriptions are provided in the `metadata.csv` file."
|
18 |
+
)
|
19 |
+
super(STTUzbekConfig, self).__init__(description=description, **kwargs)
|
20 |
+
|
21 |
+
|
22 |
+
class STTUzbek(datasets.GeneratorBasedBuilder):
|
23 |
+
DEFAULT_WRITER_BATCH_SIZE = 1000
|
24 |
+
|
25 |
+
BUILDER_CONFIGS = [
|
26 |
+
STTUzbekConfig(
|
27 |
+
name="stt_uzbek",
|
28 |
+
version=datasets.Version("1.0.0"),
|
29 |
+
),
|
30 |
+
]
|
31 |
+
|
32 |
+
def _info(self):
|
33 |
+
features = datasets.Features(
|
34 |
+
{
|
35 |
+
"file_name": datasets.Value("string"),
|
36 |
+
"audio": datasets.features.Audio(sampling_rate=48_000),
|
37 |
+
"transcription": datasets.Value("string"),
|
38 |
+
}
|
39 |
+
)
|
40 |
+
|
41 |
+
return datasets.DatasetInfo(
|
42 |
+
description=self.config.description,
|
43 |
+
features=features,
|
44 |
+
supervised_keys=None,
|
45 |
+
homepage="https://huggingface.co/datasets/Beehzod/STT_uz",
|
46 |
+
license="Apache License 2.0",
|
47 |
+
citation="""@misc{uzbek_stt_dataset,
|
48 |
+
author = {Beehzod},
|
49 |
+
title = {Uzbek Speech-to-Text Dataset},
|
50 |
+
year = {2024},
|
51 |
+
howpublished = {https://huggingface.co/datasets/Beehzod/STT_uz},
|
52 |
+
note = {Dataset for Uzbek language speech-to-text tasks.}
|
53 |
+
}""",
|
54 |
+
version=self.config.version,
|
55 |
+
)
|
56 |
+
|
57 |
+
def _split_generators(self, dl_manager):
|
58 |
+
# Adjust the paths according to your setup
|
59 |
+
wavs_dir = "audio/wavs"
|
60 |
+
uz_other_dir = "audio/uz_other_dataset"
|
61 |
+
uz_validated_dir = "audio/uz_validated_dataset"
|
62 |
+
uz_train_dir = "audio/uz_train_dataset"
|
63 |
+
metadata_file = "metadata.csv"
|
64 |
+
|
65 |
+
return [
|
66 |
+
datasets.SplitGenerator(
|
67 |
+
name=datasets.Split.TRAIN,
|
68 |
+
gen_kwargs={
|
69 |
+
"wavs_dir": wavs_dir,
|
70 |
+
"uz_other_dir": uz_other_dir,
|
71 |
+
"uz_validated_dir": uz_validated_dir,
|
72 |
+
"uz_train_dir": uz_train_dir,
|
73 |
+
"metadata_file": metadata_file,
|
74 |
+
},
|
75 |
+
),
|
76 |
+
]
|
77 |
+
|
78 |
+
def _generate_examples(self, wavs_dir, uz_other_dir, uz_validated_dir, uz_train_dir, metadata_file):
|
79 |
+
with open(metadata_file, encoding="utf-8") as f:
|
80 |
+
reader = csv.DictReader(f)
|
81 |
+
for row in tqdm(reader, desc="Processing metadata..."):
|
82 |
+
file_name = row["file_name"]
|
83 |
+
transcription = row["transcription"]
|
84 |
+
|
85 |
+
# Determine the file's location based on the path prefix
|
86 |
+
if file_name.startswith("audio/wavs"):
|
87 |
+
audio_path = os.path.join(wavs_dir, os.path.basename(file_name))
|
88 |
+
elif file_name.startswith("audio/uz_other_dataset"):
|
89 |
+
audio_path = self._extract_from_tar(uz_other_dir, file_name)
|
90 |
+
elif file_name.startswith("audio/uz_validated_dataset"):
|
91 |
+
audio_path = self._extract_from_tar(uz_validated_dir, file_name)
|
92 |
+
elif file_name.startswith("audio/uz_train_dataset"):
|
93 |
+
audio_path = self._extract_from_tar(uz_train_dir, file_name)
|
94 |
+
else:
|
95 |
+
raise ValueError(f"Unknown path prefix in file_name: {file_name}")
|
96 |
+
|
97 |
+
# Yield the example
|
98 |
+
yield file_name, {
|
99 |
+
"file_name": file_name,
|
100 |
+
"audio": audio_path,
|
101 |
+
"transcription": transcription,
|
102 |
+
}
|
103 |
+
|
104 |
+
def _extract_from_tar(self, tar_dir, file_name):
|
105 |
+
# Extract the specific file from the tar archives
|
106 |
+
for tar_file in os.listdir(tar_dir):
|
107 |
+
tar_path = os.path.join(tar_dir, tar_file)
|
108 |
+
with tarfile.open(tar_path, "r") as tar:
|
109 |
+
try:
|
110 |
+
file_path = file_name.split("/")[-1]
|
111 |
+
extracted_file = tar.extractfile(file_path)
|
112 |
+
if extracted_file:
|
113 |
+
return {"path": file_path, "bytes": extracted_file.read()}
|
114 |
+
except KeyError:
|
115 |
+
continue
|
116 |
+
raise FileNotFoundError(f"File {file_name} not found in any tar archives in {tar_dir}.")
|